diff options
| author | Billy Donahue <billy.donahue@mongodb.com> | 2019-09-10 19:03:36 +0000 |
|---|---|---|
| committer | evergreen <evergreen@mongodb.com> | 2019-09-10 19:03:36 +0000 |
| commit | 5a0f718e1309a4484580d8038016d043ef3b887f (patch) | |
| tree | 7a331d57ba6e33e61f96ed69b2aa387003a039ab /src | |
| parent | b9e29cd56ebc9aca06f68eeeda7c523d3dfd6d41 (diff) | |
| download | mongo-5a0f718e1309a4484580d8038016d043ef3b887f.tar.gz | |
SERVER-43119 FailPoint cleanup
- Don't use MONGO_INITIALIZER to declare each fail point.
We only need one init task in total: freeze and iterate the registry.
- remove MONGO_FAIL_POINT_DECLARE macro (extern)
- remove MONGO_FAIL_POINT_SHOULD_FAIL macro (FailPoint::shouldFail)
- remove MONGO_FAIL_POINT_BLOCK_IF (FailPoint::executeIf)
- remove MONGO_FAIL_POINT_BLOCK (FailPoint::execute)
- clean up FailPointRegistry and fail_point_service implementation.
Diffstat (limited to 'src')
150 files changed, 919 insertions, 978 deletions
diff --git a/src/mongo/client/dbclient_connection.cpp b/src/mongo/client/dbclient_connection.cpp index bc65eb35938..0d058396c0f 100644 --- a/src/mongo/client/dbclient_connection.cpp +++ b/src/mongo/client/dbclient_connection.cpp @@ -577,7 +577,7 @@ void DBClientConnection::say(Message& toSend, bool isRetry, string* actualServer toSend.header().setId(nextMessageId()); toSend.header().setResponseToMsgId(0); - if (!MONGO_FAIL_POINT(dbClientConnectionDisableChecksum)) { + if (!MONGO_unlikely(dbClientConnectionDisableChecksum.shouldFail())) { #ifdef MONGO_CONFIG_SSL if (!SSLPeerInfo::forSession(_session).isTLS) { OpMsg::appendChecksum(&toSend); @@ -627,7 +627,7 @@ bool DBClientConnection::call(Message& toSend, toSend.header().setId(nextMessageId()); toSend.header().setResponseToMsgId(0); - if (!MONGO_FAIL_POINT(dbClientConnectionDisableChecksum)) { + if (!MONGO_unlikely(dbClientConnectionDisableChecksum.shouldFail())) { #ifdef MONGO_CONFIG_SSL if (!SSLPeerInfo::forSession(_session).isTLS) { OpMsg::appendChecksum(&toSend); diff --git a/src/mongo/client/replica_set_monitor.cpp b/src/mongo/client/replica_set_monitor.cpp index 77d0788466a..fd79717719c 100644 --- a/src/mongo/client/replica_set_monitor.cpp +++ b/src/mongo/client/replica_set_monitor.cpp @@ -180,13 +180,12 @@ const Seconds ReplicaSetMonitor::kDefaultFindHostTimeout(15); bool ReplicaSetMonitor::useDeterministicHostSelection = false; Seconds ReplicaSetMonitor::getDefaultRefreshPeriod() { - MONGO_FAIL_POINT_BLOCK_IF(modifyReplicaSetMonitorDefaultRefreshPeriod, - data, - [&](const BSONObj& data) { return data.hasField("period"); }) { - return Seconds{data.getData().getIntField("period")}; - } - - return kDefaultRefreshPeriod; + Seconds r = kDefaultRefreshPeriod; + static constexpr auto kPeriodField = "period"_sd; + modifyReplicaSetMonitorDefaultRefreshPeriod.executeIf( + [&r](const BSONObj& data) { r = Seconds{data.getIntField(kPeriodField)}; }, + [](const BSONObj& data) { return data.hasField(kPeriodField); }); + return r; } ReplicaSetMonitor::ReplicaSetMonitor(const SetStatePtr& initialState) : _state(initialState) {} diff --git a/src/mongo/db/catalog/coll_mod.cpp b/src/mongo/db/catalog/coll_mod.cpp index a1f87acd21e..699acb8d407 100644 --- a/src/mongo/db/catalog/coll_mod.cpp +++ b/src/mongo/db/catalog/coll_mod.cpp @@ -267,7 +267,7 @@ Status _collModInternal(OperationContext* opCtx, Database* const db = autoDb.getDb(); Collection* coll = db ? db->getCollection(opCtx, nss) : nullptr; - MONGO_FAIL_POINT_PAUSE_WHILE_SET(hangAfterDatabaseLock); + hangAfterDatabaseLock.pauseWhileSet(); // May also modify a view instead of a collection. boost::optional<ViewDefinition> view; @@ -370,7 +370,7 @@ Status _collModInternal(OperationContext* opCtx, opCtx->recoveryUnit()->registerChange( std::make_unique<CollModResultChange>(oldExpireSecs, newExpireSecs, result)); - if (MONGO_FAIL_POINT(assertAfterIndexUpdate)) { + if (MONGO_unlikely(assertAfterIndexUpdate.shouldFail())) { log() << "collMod - assertAfterIndexUpdate fail point enabled."; uasserted(50970, "trigger rollback after the index update"); } diff --git a/src/mongo/db/catalog/collection_catalog_helper.cpp b/src/mongo/db/catalog/collection_catalog_helper.cpp index 620a9a2a08a..1085d7edb51 100644 --- a/src/mongo/db/catalog/collection_catalog_helper.cpp +++ b/src/mongo/db/catalog/collection_catalog_helper.cpp @@ -76,7 +76,7 @@ void forEachCollectionFromDb(OperationContext* opCtx, if (!callback(collection)) break; - MONGO_FAIL_POINT_PAUSE_WHILE_SET(hangBeforeGettingNextCollection); + hangBeforeGettingNextCollection.pauseWhileSet(); } } diff --git a/src/mongo/db/catalog/collection_impl.cpp b/src/mongo/db/catalog/collection_impl.cpp index 44cc050ce07..dae155bdb28 100644 --- a/src/mongo/db/catalog/collection_impl.cpp +++ b/src/mongo/db/catalog/collection_impl.cpp @@ -104,19 +104,21 @@ MONGO_FAIL_POINT_DEFINE(failAfterBulkLoadDocInsert); * Otherwise, the function should fail and return early with the error Status. */ Status checkFailCollectionInsertsFailPoint(const NamespaceString& ns, const BSONObj& firstDoc) { - MONGO_FAIL_POINT_BLOCK(failCollectionInserts, extraData) { - const BSONObj& data = extraData.getData(); - const auto collElem = data["collectionNS"]; - // If the failpoint specifies no collection or matches the existing one, fail. - if (!collElem || ns.ns() == collElem.str()) { + Status s = Status::OK(); + failCollectionInserts.executeIf( + [&](const BSONObj& data) { const std::string msg = str::stream() << "Failpoint (failCollectionInserts) has been enabled (" << data << "), so rejecting insert (first doc): " << firstDoc; log() << msg; - return {ErrorCodes::FailPointEnabled, msg}; - } - } - return Status::OK(); + s = {ErrorCodes::FailPointEnabled, msg}; + }, + [&](const BSONObj& data) { + // If the failpoint specifies no collection or matches the existing one, fail. + const auto collElem = data["collectionNS"]; + return !collElem || ns.ns() == collElem.str(); + }); + return s; } // Uses the collator factory to convert the BSON representation of a collator to a @@ -186,10 +188,6 @@ StatusWith<CollectionImpl::ValidationAction> _parseValidationAction(StringData n } // namespace -using std::string; -using std::unique_ptr; -using std::vector; - CollectionImpl::CollectionImpl(OperationContext* opCtx, const NamespaceString& nss, UUID uuid, @@ -361,8 +359,8 @@ Status CollectionImpl::insertDocumentsForOplog(OperationContext* opCtx, Status CollectionImpl::insertDocuments(OperationContext* opCtx, - const vector<InsertStatement>::const_iterator begin, - const vector<InsertStatement>::const_iterator end, + const std::vector<InsertStatement>::const_iterator begin, + const std::vector<InsertStatement>::const_iterator end, OpDebug* opDebug, bool fromMigrate) { @@ -401,25 +399,27 @@ Status CollectionImpl::insertDocuments(OperationContext* opCtx, opCtx->recoveryUnit()->onCommit( [this](boost::optional<Timestamp>) { notifyCappedWaitersIfNeeded(); }); - MONGO_FAIL_POINT_BLOCK(hangAfterCollectionInserts, extraData) { - const BSONObj& data = extraData.getData(); - const auto collElem = data["collectionNS"]; - const auto firstIdElem = data["first_id"]; - // If the failpoint specifies no collection or matches the existing one, hang. - if ((!collElem || _ns.ns() == collElem.str()) && - (!firstIdElem || - (begin != end && firstIdElem.type() == mongo::String && - begin->doc["_id"].str() == firstIdElem.str()))) { - string whenFirst = - firstIdElem ? (string(" when first _id is ") + firstIdElem.str()) : ""; - while (MONGO_FAIL_POINT(hangAfterCollectionInserts)) { - log() << "hangAfterCollectionInserts fail point enabled for " << _ns << whenFirst - << ". Blocking until fail point is disabled."; - mongo::sleepsecs(1); - opCtx->checkForInterrupt(); + hangAfterCollectionInserts.executeIf( + [&](const BSONObj& data) { + const auto& firstIdElem = data["first_id"]; + std::string whenFirst; + if (firstIdElem) { + whenFirst += " when first _id is "; + whenFirst += firstIdElem.str(); } - } - } + log() << "hangAfterCollectionInserts fail point enabled for " << _ns << whenFirst + << ". Blocking until fail point is disabled."; + hangAfterCollectionInserts.pauseWhileSet(opCtx); + }, + [&](const BSONObj& data) { + const auto& collElem = data["collectionNS"]; + const auto& firstIdElem = data["first_id"]; + // If the failpoint specifies no collection or matches the existing one, hang. + return (!collElem || _ns.ns() == collElem.str()) && + (!firstIdElem || + (begin != end && firstIdElem.type() == mongo::String && + begin->doc["_id"].str() == firstIdElem.str())); + }); return Status::OK(); } @@ -428,7 +428,7 @@ Status CollectionImpl::insertDocument(OperationContext* opCtx, const InsertStatement& docToInsert, OpDebug* opDebug, bool fromMigrate) { - vector<InsertStatement> docs; + std::vector<InsertStatement> docs; docs.push_back(docToInsert); return insertDocuments(opCtx, docs.begin(), docs.end(), opDebug, fromMigrate); } @@ -459,13 +459,13 @@ Status CollectionImpl::insertDocumentForBulkLoader(OperationContext* opCtx, status = onRecordInserted(loc.getValue()); - if (MONGO_FAIL_POINT(failAfterBulkLoadDocInsert)) { + if (MONGO_unlikely(failAfterBulkLoadDocInsert.shouldFail())) { log() << "Failpoint failAfterBulkLoadDocInsert enabled for " << _ns.ns() << ". Throwing WriteConflictException."; throw WriteConflictException(); } - vector<InsertStatement> inserts; + std::vector<InsertStatement> inserts; OplogSlot slot; // Fetch a new optime now, if necessary. auto replCoord = repl::ReplicationCoordinator::get(opCtx); @@ -485,8 +485,8 @@ Status CollectionImpl::insertDocumentForBulkLoader(OperationContext* opCtx, } Status CollectionImpl::_insertDocuments(OperationContext* opCtx, - const vector<InsertStatement>::const_iterator begin, - const vector<InsertStatement>::const_iterator end, + const std::vector<InsertStatement>::const_iterator begin, + const std::vector<InsertStatement>::const_iterator end, OpDebug* opDebug) { dassert(opCtx->lockState()->isCollectionLockedForMode(ns(), MODE_IX)); @@ -785,7 +785,7 @@ Status CollectionImpl::truncate(OperationContext* opCtx) { invariant(_indexCatalog->numIndexesInProgress(opCtx) == 0); // 1) store index specs - vector<BSONObj> indexSpecs; + std::vector<BSONObj> indexSpecs; { std::unique_ptr<IndexCatalog::IndexIterator> ii = _indexCatalog->getIndexIterator(opCtx, false); diff --git a/src/mongo/db/catalog/collection_validation.cpp b/src/mongo/db/catalog/collection_validation.cpp index bfd16f5e1a2..e7c0b469b71 100644 --- a/src/mongo/db/catalog/collection_validation.cpp +++ b/src/mongo/db/catalog/collection_validation.cpp @@ -565,11 +565,11 @@ Status validate(OperationContext* opCtx, // // Only useful for background validation because we hold an intent lock instead of an // exclusive lock, and thus allow concurrent operations. - if (MONGO_FAIL_POINT(pauseCollectionValidationWithLock)) { + if (MONGO_unlikely(pauseCollectionValidationWithLock.shouldFail())) { invariant(opCtx->lockState()->isCollectionLockedForMode(collection->ns(), MODE_IX)); _validationIsPausedForTest.store(true); log() << "Failpoint 'pauseCollectionValidationWithLock' activated."; - MONGO_FAIL_POINT_PAUSE_WHILE_SET(pauseCollectionValidationWithLock); + pauseCollectionValidationWithLock.pauseWhileSet(); _validationIsPausedForTest.store(false); } diff --git a/src/mongo/db/catalog/database_impl.cpp b/src/mongo/db/catalog/database_impl.cpp index c916effda65..df1f9de3931 100644 --- a/src/mongo/db/catalog/database_impl.cpp +++ b/src/mongo/db/catalog/database_impl.cpp @@ -685,7 +685,7 @@ Collection* DatabaseImpl::createCollection(OperationContext* opCtx, } } - MONGO_FAIL_POINT_PAUSE_WHILE_SET(hangBeforeLoggingCreateCollection); + hangBeforeLoggingCreateCollection.pauseWhileSet(); opCtx->getServiceContext()->getOpObserver()->onCreateCollection( opCtx, collection, nss, optionsWithUUID, fullIdIndexSpec, createOplogSlot); diff --git a/src/mongo/db/catalog/drop_collection.cpp b/src/mongo/db/catalog/drop_collection.cpp index b5f30ae2a3e..e3f9f9795ab 100644 --- a/src/mongo/db/catalog/drop_collection.cpp +++ b/src/mongo/db/catalog/drop_collection.cpp @@ -67,10 +67,10 @@ Status _dropView(OperationContext* opCtx, // Operations all lock system.views in the end to prevent deadlock. Lock::CollectionLock systemViewsLock(opCtx, db->getSystemViewsName(), MODE_X); - if (MONGO_FAIL_POINT(hangDuringDropCollection)) { + if (MONGO_unlikely(hangDuringDropCollection.shouldFail())) { log() << "hangDuringDropCollection fail point enabled. Blocking until fail point is " "disabled."; - MONGO_FAIL_POINT_PAUSE_WHILE_SET(hangDuringDropCollection); + hangDuringDropCollection.pauseWhileSet(); } AutoStatsTracker statsTracker(opCtx, @@ -108,10 +108,10 @@ Status _dropCollection(OperationContext* opCtx, return Status(ErrorCodes::NamespaceNotFound, "ns not found"); } - if (MONGO_FAIL_POINT(hangDuringDropCollection)) { + if (MONGO_unlikely(hangDuringDropCollection.shouldFail())) { log() << "hangDuringDropCollection fail point enabled. Blocking until fail point is " "disabled."; - MONGO_FAIL_POINT_PAUSE_WHILE_SET(hangDuringDropCollection); + hangDuringDropCollection.pauseWhileSet(); } AutoStatsTracker statsTracker(opCtx, @@ -156,9 +156,9 @@ Status dropCollection(OperationContext* opCtx, log() << "CMD: drop " << collectionName; } - if (MONGO_FAIL_POINT(hangDropCollectionBeforeLockAcquisition)) { + if (MONGO_unlikely(hangDropCollectionBeforeLockAcquisition.shouldFail())) { log() << "Hanging drop collection before lock acquisition while fail point is set"; - MONGO_FAIL_POINT_PAUSE_WHILE_SET(hangDropCollectionBeforeLockAcquisition); + hangDropCollectionBeforeLockAcquisition.pauseWhileSet(); } return writeConflictRetry(opCtx, "drop", collectionName.ns(), [&] { AutoGetDb autoDb(opCtx, collectionName.db(), MODE_IX); diff --git a/src/mongo/db/catalog/drop_database.cpp b/src/mongo/db/catalog/drop_database.cpp index 60b507f6d9d..d81908b9d48 100644 --- a/src/mongo/db/catalog/drop_database.cpp +++ b/src/mongo/db/catalog/drop_database.cpp @@ -85,10 +85,10 @@ void _finishDropDatabase(OperationContext* opCtx, log() << "dropDatabase " << dbName << " - dropped " << numCollections << " collection(s)"; log() << "dropDatabase " << dbName << " - finished"; - if (MONGO_FAIL_POINT(dropDatabaseHangBeforeLog)) { + if (MONGO_unlikely(dropDatabaseHangBeforeLog.shouldFail())) { log() << "dropDatabase - fail point dropDatabaseHangBeforeLog enabled. " "Blocking until fail point is disabled. "; - MONGO_FAIL_POINT_PAUSE_WHILE_SET(dropDatabaseHangBeforeLog); + dropDatabaseHangBeforeLog.pauseWhileSet(); } writeConflictRetry(opCtx, "dropDatabase_database", dbName, [&] { @@ -292,10 +292,10 @@ Status dropDatabase(OperationContext* opCtx, const std::string& dbName) { << result.duration << ". dropping database"; } - if (MONGO_FAIL_POINT(dropDatabaseHangAfterAllCollectionsDrop)) { + if (MONGO_unlikely(dropDatabaseHangAfterAllCollectionsDrop.shouldFail())) { log() << "dropDatabase - fail point dropDatabaseHangAfterAllCollectionsDrop enabled. " "Blocking until fail point is disabled. "; - MONGO_FAIL_POINT_PAUSE_WHILE_SET(dropDatabaseHangAfterAllCollectionsDrop); + dropDatabaseHangAfterAllCollectionsDrop.pauseWhileSet(); } AutoGetDb autoDB(opCtx, dbName, MODE_X); diff --git a/src/mongo/db/catalog/index_key_validate.cpp b/src/mongo/db/catalog/index_key_validate.cpp index 4b1f35ce1a6..164876fe330 100644 --- a/src/mongo/db/catalog/index_key_validate.cpp +++ b/src/mongo/db/catalog/index_key_validate.cpp @@ -503,7 +503,7 @@ Status validateIdIndexSpec(const BSONObj& indexSpec) { * expected fields are present at creation time */ Status validateIndexSpecFieldNames(const BSONObj& indexSpec) { - if (MONGO_FAIL_POINT(skipIndexCreateFieldNameValidation)) { + if (MONGO_unlikely(skipIndexCreateFieldNameValidation.shouldFail())) { return Status::OK(); } diff --git a/src/mongo/db/catalog/multi_index_block.cpp b/src/mongo/db/catalog/multi_index_block.cpp index eb65fd51f11..1eb834bd5bb 100644 --- a/src/mongo/db/catalog/multi_index_block.cpp +++ b/src/mongo/db/catalog/multi_index_block.cpp @@ -415,13 +415,16 @@ StatusWith<std::vector<BSONObj>> MultiIndexBlock::init(OperationContext* opCtx, } void failPointHangDuringBuild(FailPoint* fp, StringData where, const BSONObj& doc) { - MONGO_FAIL_POINT_BLOCK(*fp, data) { - int i = doc.getIntField("i"); - if (data.getData()["i"].numberInt() == i) { + fp->executeIf( + [&](const BSONObj& data) { + int i = doc.getIntField("i"); log() << "Hanging " << where << " index build of i=" << i; - MONGO_FAIL_POINT_PAUSE_WHILE_SET((*fp)); - } - } + fp->pauseWhileSet(); + }, + [&](const BSONObj& data) { + int i = doc.getIntField("i"); + return data["i"].numberInt() == i; + }); } Status MultiIndexBlock::insertAllDocumentsInCollection(OperationContext* opCtx, @@ -451,16 +454,16 @@ Status MultiIndexBlock::insertAllDocumentsInCollection(OperationContext* opCtx, progress.set(CurOp::get(opCtx)->setProgress_inlock(curopMessage, numRecords)); } - if (MONGO_FAIL_POINT(hangAfterSettingUpIndexBuild)) { + if (MONGO_unlikely(hangAfterSettingUpIndexBuild.shouldFail())) { // Hang the build after the BackgroundOperation and curOP info is set up. log() << "Hanging index build due to failpoint 'hangAfterSettingUpIndexBuild'"; - MONGO_FAIL_POINT_PAUSE_WHILE_SET(hangAfterSettingUpIndexBuild); + hangAfterSettingUpIndexBuild.pauseWhileSet(); } - if (MONGO_FAIL_POINT(hangAndThenFailIndexBuild)) { + if (MONGO_unlikely(hangAndThenFailIndexBuild.shouldFail())) { // Hang the build after the BackgroundOperation and curOP info is set up. log() << "Hanging index build due to failpoint 'hangAndThenFailIndexBuild'"; - MONGO_FAIL_POINT_PAUSE_WHILE_SET(hangAndThenFailIndexBuild); + hangAndThenFailIndexBuild.pauseWhileSet(); return {ErrorCodes::InternalError, "Failed index build because of failpoint 'hangAndThenFailIndexBuild'"}; } @@ -494,7 +497,7 @@ Status MultiIndexBlock::insertAllDocumentsInCollection(OperationContext* opCtx, int retries = 0; // non-zero when retrying our last document. while (retries || (PlanExecutor::ADVANCED == (state = exec->getNextSnapshotted(&objToIndex, &loc))) || - MONGO_FAIL_POINT(hangAfterStartingIndexBuild)) { + MONGO_unlikely(hangAfterStartingIndexBuild.shouldFail())) { try { auto interruptStatus = opCtx->checkForInterruptNoAssert(); if (!interruptStatus.isOK()) @@ -565,7 +568,7 @@ Status MultiIndexBlock::insertAllDocumentsInCollection(OperationContext* opCtx, return exec->getMemberObjectStatus(objToIndex.value()); } - if (MONGO_FAIL_POINT(leaveIndexBuildUnfinishedForShutdown)) { + if (MONGO_unlikely(leaveIndexBuildUnfinishedForShutdown.shouldFail())) { log() << "Index build interrupted due to 'leaveIndexBuildUnfinishedForShutdown' failpoint. " "Mimicing shutdown error code."; return Status( @@ -573,14 +576,14 @@ Status MultiIndexBlock::insertAllDocumentsInCollection(OperationContext* opCtx, "background index build interrupted due to failpoint. returning a shutdown error."); } - if (MONGO_FAIL_POINT(hangAfterStartingIndexBuildUnlocked)) { + if (MONGO_unlikely(hangAfterStartingIndexBuildUnlocked.shouldFail())) { // Unlock before hanging so replication recognizes we've completed. Locker::LockSnapshot lockInfo; invariant(opCtx->lockState()->saveLockStateAndUnlock(&lockInfo)); log() << "Hanging index build with no locks due to " "'hangAfterStartingIndexBuildUnlocked' failpoint"; - MONGO_FAIL_POINT_PAUSE_WHILE_SET(hangAfterStartingIndexBuildUnlocked); + hangAfterStartingIndexBuildUnlocked.pauseWhileSet(); if (isBackgroundBuilding()) { opCtx->lockState()->restoreLockState(opCtx, lockInfo); diff --git a/src/mongo/db/catalog/multi_index_block.h b/src/mongo/db/catalog/multi_index_block.h index 167b2fda9c3..69aa9cd4a34 100644 --- a/src/mongo/db/catalog/multi_index_block.h +++ b/src/mongo/db/catalog/multi_index_block.h @@ -51,7 +51,7 @@ namespace mongo { -MONGO_FAIL_POINT_DECLARE(leaveIndexBuildUnfinishedForShutdown); +extern FailPoint leaveIndexBuildUnfinishedForShutdown; class Collection; class MatchExpression; diff --git a/src/mongo/db/catalog/rename_collection.cpp b/src/mongo/db/catalog/rename_collection.cpp index 428cef3df58..1fcd21bd593 100644 --- a/src/mongo/db/catalog/rename_collection.cpp +++ b/src/mongo/db/catalog/rename_collection.cpp @@ -102,7 +102,7 @@ Status checkSourceAndTargetNamespaces(OperationContext* opCtx, << target); // TODO: SERVER-42638 Replace checks of cm() with cm()->distributionMode() == sharded - if (!MONGO_FAIL_POINT(useRenameCollectionPathThroughConfigsvr)) { + if (!MONGO_unlikely(useRenameCollectionPathThroughConfigsvr.shouldFail())) { if (isCollectionSharded(opCtx, source)) return {ErrorCodes::IllegalOperation, "source namespace cannot be sharded"}; } @@ -470,7 +470,7 @@ Status renameBetweenDBs(OperationContext* opCtx, } // TODO: SERVER-42638 Replace checks of cm() with cm()->distributionMode() == sharded - if (!MONGO_FAIL_POINT(useRenameCollectionPathThroughConfigsvr)) { + if (!MONGO_unlikely(useRenameCollectionPathThroughConfigsvr.shouldFail())) { if (isCollectionSharded(opCtx, source)) return {ErrorCodes::IllegalOperation, "source namespace cannot be sharded"}; } @@ -675,7 +675,7 @@ Status renameBetweenDBs(OperationContext* opCtx, opCtx, "retryRestoreCursor", ns, [&cursor] { cursor->restore(); }); }); // Used to make sure that a WCE can be handled by this logic without data loss. - if (MONGO_FAIL_POINT(writeConflictInRenameCollCopyToTmp)) { + if (MONGO_unlikely(writeConflictInRenameCollCopyToTmp.shouldFail())) { throw WriteConflictException(); } wunit.commit(); diff --git a/src/mongo/db/catalog/throttle_cursor.cpp b/src/mongo/db/catalog/throttle_cursor.cpp index 8e2fb2f9745..1cfd22c23f7 100644 --- a/src/mongo/db/catalog/throttle_cursor.cpp +++ b/src/mongo/db/catalog/throttle_cursor.cpp @@ -118,7 +118,7 @@ void DataThrottle::awaitIfNeeded(OperationContext* opCtx, const int64_t dataSize _bytesProcessed = 0; } - _bytesProcessed += MONGO_FAIL_POINT(fixedCursorDataSizeOf512KBForDataThrottle) + _bytesProcessed += MONGO_unlikely(fixedCursorDataSizeOf512KBForDataThrottle.shouldFail()) ? /*512KB*/ 1 * 1024 * 512 : dataSize; diff --git a/src/mongo/db/catalog_raii.cpp b/src/mongo/db/catalog_raii.cpp index a9b3b49ac28..ce29a2c495b 100644 --- a/src/mongo/db/catalog_raii.cpp +++ b/src/mongo/db/catalog_raii.cpp @@ -73,10 +73,8 @@ AutoGetCollection::AutoGetCollection(OperationContext* opCtx, _resolvedNss = CollectionCatalog::get(opCtx).resolveNamespaceStringOrUUID(nsOrUUID); // Wait for a configured amount of time after acquiring locks if the failpoint is enabled - MONGO_FAIL_POINT_BLOCK(setAutoGetCollectionWait, customWait) { - const BSONObj& data = customWait.getData(); - sleepFor(Milliseconds(data["waitForMillis"].numberInt())); - } + setAutoGetCollectionWait.execute( + [&](const BSONObj& data) { sleepFor(Milliseconds(data["waitForMillis"].numberInt())); }); Database* const db = _autoDb.getDb(); invariant(!nsOrUUID.uuid() || db, diff --git a/src/mongo/db/cloner.cpp b/src/mongo/db/cloner.cpp index 2a9c6d70b24..30402ec2552 100644 --- a/src/mongo/db/cloner.cpp +++ b/src/mongo/db/cloner.cpp @@ -533,7 +533,7 @@ Status Cloner::createCollectionsForDb( auto collCount = 0; for (auto&& params : createCollectionParams) { - if (MONGO_FAIL_POINT(movePrimaryFailPoint) && collCount > 0) { + if (MONGO_unlikely(movePrimaryFailPoint.shouldFail()) && collCount > 0) { return Status(ErrorCodes::CommandFailed, "movePrimary failed due to failpoint"); } collCount++; diff --git a/src/mongo/db/commands.cpp b/src/mongo/db/commands.cpp index 031836713a8..7ca2cec1443 100644 --- a/src/mongo/db/commands.cpp +++ b/src/mongo/db/commands.cpp @@ -506,34 +506,35 @@ bool CommandHelpers::shouldActivateFailCommandFailPoint(const BSONObj& data, void CommandHelpers::evaluateFailCommandFailPoint(OperationContext* opCtx, StringData commandName, const NamespaceString& nss) { - bool closeConnection, hasErrorCode; + bool closeConnection; + bool hasErrorCode; long long errorCode; + failCommand.executeIf( + [&](const BSONObj&) { + if (closeConnection) { + opCtx->getClient()->session()->end(); + log() << "Failing command '" << commandName + << "' via 'failCommand' failpoint. Action: closing connection."; + uasserted(50985, "Failing command due to 'failCommand' failpoint"); + } - MONGO_FAIL_POINT_BLOCK_IF(failCommand, data, [&](const BSONObj& data) { - closeConnection = data.hasField("closeConnection") && - bsonExtractBooleanField(data, "closeConnection", &closeConnection).isOK() && - closeConnection; - hasErrorCode = data.hasField("errorCode") && - bsonExtractIntegerField(data, "errorCode", &errorCode).isOK(); - - return shouldActivateFailCommandFailPoint(data, commandName, opCtx->getClient(), nss) && - (closeConnection || hasErrorCode); - }) { - if (closeConnection) { - opCtx->getClient()->session()->end(); - log() << "Failing command '" << commandName - << "' via 'failCommand' failpoint. Action: closing connection."; - uasserted(50985, "Failing command due to 'failCommand' failpoint"); - } - - if (hasErrorCode) { - log() << "Failing command '" << commandName - << "' via 'failCommand' failpoint. Action: returning error code " << errorCode - << "."; - uasserted(ErrorCodes::Error(errorCode), - "Failing command due to 'failCommand' failpoint"); - } - } + if (hasErrorCode) { + log() << "Failing command '" << commandName + << "' via 'failCommand' failpoint. Action: returning error code " << errorCode + << "."; + uasserted(ErrorCodes::Error(errorCode), + "Failing command due to 'failCommand' failpoint"); + } + }, + [&](const BSONObj& data) { + closeConnection = data.hasField("closeConnection") && + bsonExtractBooleanField(data, "closeConnection", &closeConnection).isOK() && + closeConnection; + hasErrorCode = data.hasField("errorCode") && + bsonExtractIntegerField(data, "errorCode", &errorCode).isOK(); + return shouldActivateFailCommandFailPoint(data, commandName, opCtx->getClient(), nss) && + (closeConnection || hasErrorCode); + }); } void CommandHelpers::handleMarkKillOnClientDisconnect(OperationContext* opCtx, @@ -546,16 +547,13 @@ void CommandHelpers::handleMarkKillOnClientDisconnect(OperationContext* opCtx, opCtx->markKillOnClientDisconnect(); } - MONGO_FAIL_POINT_BLOCK_IF( - waitInCommandMarkKillOnClientDisconnect, options, [&](const BSONObj& obj) { - const auto& clientMetadata = + waitInCommandMarkKillOnClientDisconnect.executeIf( + [&](const BSONObj&) { waitInCommandMarkKillOnClientDisconnect.pauseWhileSet(opCtx); }, + [&](const BSONObj& obj) { + const auto& md = ClientMetadataIsMasterState::get(opCtx->getClient()).getClientMetadata(); - - return clientMetadata && (clientMetadata->getApplicationName() == obj["appName"].str()); - }) { - MONGO_FAIL_POINT_PAUSE_WHILE_SET_OR_INTERRUPTED(opCtx, - waitInCommandMarkKillOnClientDisconnect); - } + return md && (md->getApplicationName() == obj["appName"].str()); + }); } ////////////////////////////////////////////////////////////// diff --git a/src/mongo/db/commands.h b/src/mongo/db/commands.h index f355df6289c..e944f7c238f 100644 --- a/src/mongo/db/commands.h +++ b/src/mongo/db/commands.h @@ -54,8 +54,8 @@ namespace mongo { -MONGO_FAIL_POINT_DECLARE(failCommand); -MONGO_FAIL_POINT_DECLARE(waitInCommandMarkKillOnClientDisconnect); +extern FailPoint failCommand; +extern FailPoint waitInCommandMarkKillOnClientDisconnect; class Command; class CommandInvocation; diff --git a/src/mongo/db/commands/create_indexes.cpp b/src/mongo/db/commands/create_indexes.cpp index 36befb7255e..9ccb0c4af79 100644 --- a/src/mongo/db/commands/create_indexes.cpp +++ b/src/mongo/db/commands/create_indexes.cpp @@ -487,7 +487,7 @@ bool runCreateIndexes(OperationContext* opCtx, // The 'indexer' can throw, so ensure the build cleanup occurs. ON_BLOCK_EXIT([&] { opCtx->recoveryUnit()->abandonSnapshot(); - if (MONGO_FAIL_POINT(leaveIndexBuildUnfinishedForShutdown)) { + if (MONGO_unlikely(leaveIndexBuildUnfinishedForShutdown.shouldFail())) { // Set a flag to leave the persisted index build state intact when cleanUpAfterBuild() // is called below. The index build will be found on server startup. // @@ -533,9 +533,9 @@ bool runCreateIndexes(OperationContext* opCtx, uassertStatusOK(indexer.insertAllDocumentsInCollection(opCtx, collection)); } - if (MONGO_FAIL_POINT(hangAfterIndexBuildDumpsInsertsFromBulk)) { + if (MONGO_unlikely(hangAfterIndexBuildDumpsInsertsFromBulk.shouldFail())) { log() << "Hanging after dumping inserts from bulk builder"; - MONGO_FAIL_POINT_PAUSE_WHILE_SET(hangAfterIndexBuildDumpsInsertsFromBulk); + hangAfterIndexBuildDumpsInsertsFromBulk.pauseWhileSet(); } // Perform the first drain while holding an intent lock. @@ -554,9 +554,9 @@ bool runCreateIndexes(OperationContext* opCtx, uassertStatusOK(indexer.drainBackgroundWrites(opCtx)); } - if (MONGO_FAIL_POINT(hangAfterIndexBuildFirstDrain)) { + if (MONGO_unlikely(hangAfterIndexBuildFirstDrain.shouldFail())) { log() << "Hanging after index build first drain"; - MONGO_FAIL_POINT_PAUSE_WHILE_SET(hangAfterIndexBuildFirstDrain); + hangAfterIndexBuildFirstDrain.pauseWhileSet(); } // Perform the second drain while stopping writes on the collection. @@ -575,9 +575,9 @@ bool runCreateIndexes(OperationContext* opCtx, uassertStatusOK(indexer.drainBackgroundWrites(opCtx)); } - if (MONGO_FAIL_POINT(hangAfterIndexBuildSecondDrain)) { + if (MONGO_unlikely(hangAfterIndexBuildSecondDrain.shouldFail())) { log() << "Hanging after index build second drain"; - MONGO_FAIL_POINT_PAUSE_WHILE_SET(hangAfterIndexBuildSecondDrain); + hangAfterIndexBuildSecondDrain.pauseWhileSet(); } // Need to get exclusive collection lock back to complete the index build. diff --git a/src/mongo/db/commands/drop_indexes.cpp b/src/mongo/db/commands/drop_indexes.cpp index 3aba920862c..963a7aa7f83 100644 --- a/src/mongo/db/commands/drop_indexes.cpp +++ b/src/mongo/db/commands/drop_indexes.cpp @@ -220,7 +220,7 @@ public: }); } - if (MONGO_FAIL_POINT(reIndexCrashAfterDrop)) { + if (MONGO_unlikely(reIndexCrashAfterDrop.shouldFail())) { log() << "exiting because 'reIndexCrashAfterDrop' fail point was set"; quickExit(EXIT_ABRUPT); } diff --git a/src/mongo/db/commands/feature_compatibility_version.cpp b/src/mongo/db/commands/feature_compatibility_version.cpp index 0a7482078b4..8e7cf3b8d98 100644 --- a/src/mongo/db/commands/feature_compatibility_version.cpp +++ b/src/mongo/db/commands/feature_compatibility_version.cpp @@ -177,12 +177,11 @@ void FeatureCompatibilityVersion::onInsertOrUpdate(OperationContext* opCtx, cons } if (newVersion != ServerGlobalParams::FeatureCompatibility::Version::kFullyUpgradedTo44) { - if (MONGO_FAIL_POINT(hangBeforeAbortingRunningTransactionsOnFCVDowngrade)) { + if (MONGO_unlikely(hangBeforeAbortingRunningTransactionsOnFCVDowngrade.shouldFail())) { log() << "featureCompatibilityVersion - " "hangBeforeAbortingRunningTransactionsOnFCVDowngrade fail point enabled. " "Blocking until fail point is disabled."; - MONGO_FAIL_POINT_PAUSE_WHILE_SET( - hangBeforeAbortingRunningTransactionsOnFCVDowngrade); + hangBeforeAbortingRunningTransactionsOnFCVDowngrade.pauseWhileSet(); } // Abort all open transactions when downgrading the featureCompatibilityVersion. SessionKiller::Matcher matcherAllSessions( diff --git a/src/mongo/db/commands/generic_servers.cpp b/src/mongo/db/commands/generic_servers.cpp index 9c0a0ad661f..5fe6e5b56e9 100644 --- a/src/mongo/db/commands/generic_servers.cpp +++ b/src/mongo/db/commands/generic_servers.cpp @@ -332,13 +332,12 @@ void CmdShutdown::shutdownHelper(const BSONObj& cmdObj) { ShutdownTaskArgs shutdownArgs; shutdownArgs.isUserInitiated = true; - MONGO_FAIL_POINT_BLOCK(crashOnShutdown, crashBlock) { - const std::string crashHow = crashBlock.getData()["how"].str(); - if (crashHow == "fault") { + crashOnShutdown.execute([&](const BSONObj& data) { + if (data["how"].str() == "fault") { ++*illegalAddress; } ::abort(); - } + }); log() << "terminating, shutdown command received " << cmdObj; @@ -352,11 +351,10 @@ void CmdShutdown::shutdownHelper(const BSONObj& cmdObj) { // The ServiceMain thread will quit for us so just sleep until it does. while (true) sleepsecs(60); // Loop forever - } else -#endif - { - shutdown(EXIT_CLEAN, shutdownArgs); // this never returns + return; } +#endif + shutdown(EXIT_CLEAN, shutdownArgs); // this never returns } } // namespace mongo diff --git a/src/mongo/db/commands/getmore_cmd.cpp b/src/mongo/db/commands/getmore_cmd.cpp index c146024162f..cceb0f15fa8 100644 --- a/src/mongo/db/commands/getmore_cmd.cpp +++ b/src/mongo/db/commands/getmore_cmd.cpp @@ -385,11 +385,10 @@ public: invariant(cursorPin->lockPolicy() == ClientCursorParams::LockPolicy::kLockExternally); - if (MONGO_FAIL_POINT(GetMoreHangBeforeReadLock)) { + if (MONGO_unlikely(GetMoreHangBeforeReadLock.shouldFail())) { log() << "GetMoreHangBeforeReadLock fail point enabled. Blocking until fail " "point is disabled."; - MONGO_FAIL_POINT_PAUSE_WHILE_SET_OR_INTERRUPTED(opCtx, - GetMoreHangBeforeReadLock); + GetMoreHangBeforeReadLock.pauseWhileSet(opCtx); } // Lock the backing collection by using the executor's namespace. Note that it may @@ -445,7 +444,7 @@ public: validateLSID(opCtx, _request, cursorPin.getCursor()); validateTxnNumber(opCtx, _request, cursorPin.getCursor()); - if (_request.nss.isOplog() && MONGO_FAIL_POINT(rsStopGetMoreCmd)) { + if (_request.nss.isOplog() && MONGO_unlikely(rsStopGetMoreCmd.shouldFail())) { uasserted(ErrorCodes::CommandFailed, str::stream() << "getMore on " << _request.nss.ns() << " rejected due to active fail point rsStopGetMoreCmd"); @@ -475,7 +474,7 @@ public: readLock.reset(); readLock.emplace(opCtx, _request.nss); }; - if (MONGO_FAIL_POINT(waitAfterPinningCursorBeforeGetMoreBatch)) { + if (MONGO_unlikely(waitAfterPinningCursorBeforeGetMoreBatch.shouldFail())) { CurOpFailpointHelpers::waitWhileFailPointEnabled( &waitAfterPinningCursorBeforeGetMoreBatch, opCtx, @@ -486,7 +485,7 @@ public: } const bool disableAwaitDataFailpointActive = - MONGO_FAIL_POINT(disableAwaitDataForGetMoreCmd); + MONGO_unlikely(disableAwaitDataForGetMoreCmd.shouldFail()); // Inherit properties like readConcern and maxTimeMS from our originating cursor. setUpOperationContextStateForGetMore( @@ -553,8 +552,8 @@ public: dropAndReacquireReadLock(); exec->restoreState(); }; - MONGO_FAIL_POINT_BLOCK(waitWithPinnedCursorDuringGetMoreBatch, options) { - const BSONObj& data = options.getData(); + + waitWithPinnedCursorDuringGetMoreBatch.execute([&](const BSONObj& data) { CurOpFailpointHelpers::waitWhileFailPointEnabled( &waitWithPinnedCursorDuringGetMoreBatch, opCtx, @@ -564,7 +563,7 @@ public: : saveAndRestoreStateWithReadLockReacquisition, false, _request.nss); - } + }); uassertStatusOK(generateBatch( opCtx, cursorPin.getCursor(), _request, &nextBatch, &state, &numResults)); @@ -647,7 +646,7 @@ public: // If the 'waitBeforeUnpinningOrDeletingCursorAfterGetMoreBatch' failpoint is active, we // set the 'msg' field of this operation's CurOp to signal that we've hit this point and // then spin until the failpoint is released. - if (MONGO_FAIL_POINT(waitBeforeUnpinningOrDeletingCursorAfterGetMoreBatch)) { + if (MONGO_unlikely(waitBeforeUnpinningOrDeletingCursorAfterGetMoreBatch.shouldFail())) { CurOpFailpointHelpers::waitWhileFailPointEnabled( &waitBeforeUnpinningOrDeletingCursorAfterGetMoreBatch, opCtx, @@ -666,7 +665,7 @@ public: return AllowedOnSecondary::kAlways; } - ReadWriteType getReadWriteType() const { + ReadWriteType getReadWriteType() const override { return ReadWriteType::kRead; } diff --git a/src/mongo/db/commands/txn_cmds.cpp b/src/mongo/db/commands/txn_cmds.cpp index 8a0fe30e9ec..4575eef8a62 100644 --- a/src/mongo/db/commands/txn_cmds.cpp +++ b/src/mongo/db/commands/txn_cmds.cpp @@ -104,7 +104,8 @@ public: // commit oplog entry. auto& replClient = repl::ReplClientInfo::forClient(opCtx->getClient()); replClient.setLastOpToSystemLastOpTime(opCtx); - if (MONGO_FAIL_POINT(participantReturnNetworkErrorForCommitAfterExecutingCommitLogic)) { + if (MONGO_unlikely( + participantReturnNetworkErrorForCommitAfterExecutingCommitLogic.shouldFail())) { uasserted(ErrorCodes::HostUnreachable, "returning network error because failpoint is on"); } @@ -134,7 +135,8 @@ public: txnParticipant.commitUnpreparedTransaction(opCtx); } - if (MONGO_FAIL_POINT(participantReturnNetworkErrorForCommitAfterExecutingCommitLogic)) { + if (MONGO_unlikely( + participantReturnNetworkErrorForCommitAfterExecutingCommitLogic.shouldFail())) { uasserted(ErrorCodes::HostUnreachable, "returning network error because failpoint is on"); } @@ -189,7 +191,7 @@ public: CurOpFailpointHelpers::waitWhileFailPointEnabled( &hangBeforeAbortingTxn, opCtx, "hangBeforeAbortingTxn"); - if (!MONGO_FAIL_POINT(dontRemoveTxnCoordinatorOnAbort) && + if (!MONGO_unlikely(dontRemoveTxnCoordinatorOnAbort.shouldFail()) && (ShardingState::get(opCtx)->canAcceptShardedCommands().isOK() || serverGlobalParams.clusterRole == ClusterRole::ConfigServer)) { TransactionCoordinatorService::get(opCtx)->cancelIfCommitNotYetStarted( @@ -198,7 +200,8 @@ public: txnParticipant.abortTransaction(opCtx); - if (MONGO_FAIL_POINT(participantReturnNetworkErrorForAbortAfterExecutingAbortLogic)) { + if (MONGO_unlikely( + participantReturnNetworkErrorForAbortAfterExecutingAbortLogic.shouldFail())) { uasserted(ErrorCodes::HostUnreachable, "returning network error because failpoint is on"); } diff --git a/src/mongo/db/commands/validate.cpp b/src/mongo/db/commands/validate.cpp index 43c5fcae3b0..63ed7581a3c 100644 --- a/src/mongo/db/commands/validate.cpp +++ b/src/mongo/db/commands/validate.cpp @@ -109,7 +109,7 @@ public: const std::string& dbname, const BSONObj& cmdObj, BSONObjBuilder& result) { - if (MONGO_FAIL_POINT(validateCmdCollectionNotValid)) { + if (MONGO_unlikely(validateCmdCollectionNotValid.shouldFail())) { result.appendBool("valid", false); return true; } diff --git a/src/mongo/db/concurrency/lock_state.cpp b/src/mongo/db/concurrency/lock_state.cpp index 3a162b425b0..fd840437c89 100644 --- a/src/mongo/db/concurrency/lock_state.cpp +++ b/src/mongo/db/concurrency/lock_state.cpp @@ -895,7 +895,8 @@ void LockerImpl::_lockComplete(OperationContext* opCtx, // This failpoint is used to time out non-intent locks if they cannot be granted immediately. // Testing-only. - if (!_uninterruptibleLocksRequested && MONGO_FAIL_POINT(failNonIntentLocksIfWaitNeeded)) { + if (!_uninterruptibleLocksRequested && + MONGO_unlikely(failNonIntentLocksIfWaitNeeded.shouldFail())) { uassert(ErrorCodes::LockTimeout, str::stream() << "Cannot immediately acquire lock '" << resId.toString() << "'. Timing out due to failpoint.", @@ -991,7 +992,7 @@ void LockerImpl::getFlowControlTicket(OperationContext* opCtx, LockMode lockMode LockResult LockerImpl::lockRSTLBegin(OperationContext* opCtx, LockMode mode) { bool testOnly = false; - if (MONGO_FAIL_POINT(enableTestOnlyFlagforRSTL)) { + if (MONGO_unlikely(enableTestOnlyFlagforRSTL.shouldFail())) { testOnly = true; } diff --git a/src/mongo/db/concurrency/write_conflict_exception.h b/src/mongo/db/concurrency/write_conflict_exception.h index 71f38b34670..84089ae86f6 100644 --- a/src/mongo/db/concurrency/write_conflict_exception.h +++ b/src/mongo/db/concurrency/write_conflict_exception.h @@ -38,7 +38,7 @@ namespace mongo { -MONGO_FAIL_POINT_DECLARE(skipWriteConflictRetries); +extern FailPoint skipWriteConflictRetries; /** * This is thrown if during a write, two or more operations conflict with each other. @@ -83,7 +83,8 @@ auto writeConflictRetry(OperationContext* opCtx, StringData opStr, StringData ns invariant(opCtx->lockState()); invariant(opCtx->recoveryUnit()); - if (opCtx->lockState()->inAWriteUnitOfWork() || MONGO_FAIL_POINT(skipWriteConflictRetries)) { + if (opCtx->lockState()->inAWriteUnitOfWork() || + MONGO_unlikely(skipWriteConflictRetries.shouldFail())) { return f(); } diff --git a/src/mongo/db/curop.cpp b/src/mongo/db/curop.cpp index da117a244b4..de804da0983 100644 --- a/src/mongo/db/curop.cpp +++ b/src/mongo/db/curop.cpp @@ -247,7 +247,7 @@ void CurOp::reportCurrentOpForClient(OperationContext* opCtx, bool backtraceMode, BSONObjBuilder* infoBuilder) { invariant(client); - if (MONGO_FAIL_POINT(keepDiagnosticCaptureOnFailedLock)) { + if (MONGO_unlikely(keepDiagnosticCaptureOnFailedLock.shouldFail())) { gHangLock.lock.lock(); try { stdx::lock_guard testLock(gHangLock.mutex); @@ -336,7 +336,7 @@ void CurOp::reportCurrentOpForClient(OperationContext* opCtx, } } - if (MONGO_FAIL_POINT(keepDiagnosticCaptureOnFailedLock)) { + if (MONGO_unlikely(keepDiagnosticCaptureOnFailedLock.shouldFail())) { gHangLock.lock.unlock(); } } diff --git a/src/mongo/db/curop_failpoint_helpers.cpp b/src/mongo/db/curop_failpoint_helpers.cpp index 6afbfb05be5..5f2a4190de7 100644 --- a/src/mongo/db/curop_failpoint_helpers.cpp +++ b/src/mongo/db/curop_failpoint_helpers.cpp @@ -47,42 +47,44 @@ std::string CurOpFailpointHelpers::updateCurOpMsg(OperationContext* opCtx, void CurOpFailpointHelpers::waitWhileFailPointEnabled(FailPoint* failPoint, OperationContext* opCtx, const std::string& curOpMsg, - const std::function<void(void)>& whileWaiting, + const std::function<void()>& whileWaiting, bool checkForInterrupt, boost::optional<NamespaceString> nss) { - invariant(failPoint); - MONGO_FAIL_POINT_BLOCK((*failPoint), options) { - const BSONObj& data = options.getData(); - StringData fpNss = data.getStringField("nss"); - if (nss && !fpNss.empty() && fpNss != nss.get().toString()) { - return; - } + failPoint->executeIf( + [&](const BSONObj& data) { + auto origCurOpMsg = updateCurOpMsg(opCtx, curOpMsg); - auto origCurOpMsg = updateCurOpMsg(opCtx, curOpMsg); + const bool shouldCheckForInterrupt = + checkForInterrupt || data["shouldCheckForInterrupt"].booleanSafe(); + const bool shouldContinueOnInterrupt = data["shouldContinueOnInterrupt"].booleanSafe(); + while (MONGO_unlikely(failPoint->shouldFail())) { + sleepFor(Milliseconds(10)); + if (whileWaiting) { + whileWaiting(); + } - const bool shouldCheckForInterrupt = - checkForInterrupt || data["shouldCheckForInterrupt"].booleanSafe(); - const bool shouldContinueOnInterrupt = data["shouldContinueOnInterrupt"].booleanSafe(); - while (MONGO_FAIL_POINT((*failPoint))) { - sleepFor(Milliseconds(10)); - if (whileWaiting) { - whileWaiting(); + // Check for interrupt so that an operation can be killed while waiting for the + // failpoint to be disabled, if the failpoint is configured to be interruptible. + // + // For shouldContinueOnInterrupt, an interrupt merely allows the code to continue + // past the failpoint; it is up to the code under test to actually check for + // interrupt. + if (shouldContinueOnInterrupt) { + if (!opCtx->checkForInterruptNoAssert().isOK()) + break; + } else if (shouldCheckForInterrupt) { + opCtx->checkForInterrupt(); + } } - - // Check for interrupt so that an operation can be killed while waiting for the - // failpoint to be disabled, if the failpoint is configured to be interruptible. - // - // For shouldContinueOnInterrupt, an interrupt merely allows the code to continue past - // the failpoint; it is up to the code under test to actually check for interrupt. - if (shouldContinueOnInterrupt) { - if (!opCtx->checkForInterruptNoAssert().isOK()) - break; - } else if (shouldCheckForInterrupt) { - opCtx->checkForInterrupt(); + updateCurOpMsg(opCtx, origCurOpMsg); + }, + [&](const BSONObj& data) { + StringData fpNss = data.getStringField("nss"); + if (nss && !fpNss.empty() && fpNss != nss.get().toString()) { + return false; } - } - updateCurOpMsg(opCtx, origCurOpMsg); - } + return true; + }); } } // namespace mongo diff --git a/src/mongo/db/curop_failpoint_helpers.h b/src/mongo/db/curop_failpoint_helpers.h index a1143805951..8e2bd1bc08f 100644 --- a/src/mongo/db/curop_failpoint_helpers.h +++ b/src/mongo/db/curop_failpoint_helpers.h @@ -60,7 +60,7 @@ public: static void waitWhileFailPointEnabled(FailPoint* failPoint, OperationContext* opCtx, const std::string& curOpMsg, - const std::function<void(void)>& whileWaiting = nullptr, + const std::function<void()>& whileWaiting = nullptr, bool checkForInterrupt = false, boost::optional<NamespaceString> nss = boost::none); }; diff --git a/src/mongo/db/db.cpp b/src/mongo/db/db.cpp index 1f76b6e6d03..f0778289bd8 100644 --- a/src/mongo/db/db.cpp +++ b/src/mongo/db/db.cpp @@ -669,7 +669,7 @@ ExitCode _initAndListen(int listenPort) { } #endif - if (MONGO_FAIL_POINT(shutdownAtStartup)) { + if (MONGO_unlikely(shutdownAtStartup.shouldFail())) { log() << "starting clean exit via failpoint"; exitCleanly(EXIT_CLEAN); } diff --git a/src/mongo/db/exec/update_stage.cpp b/src/mongo/db/exec/update_stage.cpp index 4da1d88b854..a9bddf56266 100644 --- a/src/mongo/db/exec/update_stage.cpp +++ b/src/mongo/db/exec/update_stage.cpp @@ -607,7 +607,7 @@ void UpdateStage::doInsert() { } } - if (MONGO_FAIL_POINT(hangBeforeUpsertPerformsInsert)) { + if (MONGO_unlikely(hangBeforeUpsertPerformsInsert.shouldFail())) { CurOpFailpointHelpers::waitWhileFailPointEnabled( &hangBeforeUpsertPerformsInsert, getOpCtx(), "hangBeforeUpsertPerformsInsert"); } @@ -972,10 +972,9 @@ bool UpdateStage::checkUpdateChangesShardKeyFields(ScopedCollectionMetadata meta getOpCtx()->getTxnNumber() || !getOpCtx()->writesAreReplicated()); if (!metadata->keyBelongsToMe(newShardKey)) { - if (MONGO_FAIL_POINT(hangBeforeThrowWouldChangeOwningShard)) { + if (MONGO_unlikely(hangBeforeThrowWouldChangeOwningShard.shouldFail())) { log() << "Hit hangBeforeThrowWouldChangeOwningShard failpoint"; - MONGO_FAIL_POINT_PAUSE_WHILE_SET_OR_INTERRUPTED(getOpCtx(), - hangBeforeThrowWouldChangeOwningShard); + hangBeforeThrowWouldChangeOwningShard.pauseWhileSet(getOpCtx()); } uasserted(WouldChangeOwningShardInfo(oldObj.value(), newObj, false /* upsert */), diff --git a/src/mongo/db/index/index_build_interceptor.cpp b/src/mongo/db/index/index_build_interceptor.cpp index a6a53894d1b..c55a569442f 100644 --- a/src/mongo/db/index/index_build_interceptor.cpp +++ b/src/mongo/db/index/index_build_interceptor.cpp @@ -331,13 +331,14 @@ void IndexBuildInterceptor::_tryYield(OperationContext* opCtx) { // Track the number of yields in CurOp. CurOp::get(opCtx)->yielded(); - MONGO_FAIL_POINT_BLOCK(hangDuringIndexBuildDrainYield, config) { - StringData ns{config.getData().getStringField("namespace")}; - if (ns == _indexCatalogEntry->ns().ns()) { + hangDuringIndexBuildDrainYield.executeIf( + [&](auto&&) { log() << "Hanging index build during drain yield"; - MONGO_FAIL_POINT_PAUSE_WHILE_SET(hangDuringIndexBuildDrainYield); - } - } + hangDuringIndexBuildDrainYield.pauseWhileSet(); + }, + [&](auto&& config) { + return config.getStringField("namespace") == _indexCatalogEntry->ns().ns(); + }); locker->restoreLockState(opCtx, snapshot); } diff --git a/src/mongo/db/index_builds_coordinator.cpp b/src/mongo/db/index_builds_coordinator.cpp index f6e0edfdddb..2bbafd3cac5 100644 --- a/src/mongo/db/index_builds_coordinator.cpp +++ b/src/mongo/db/index_builds_coordinator.cpp @@ -894,9 +894,9 @@ void IndexBuildsCoordinator::_buildIndex( _indexBuildsManager.startBuildingIndex(opCtx, collection, replState->buildUUID)); } - if (MONGO_FAIL_POINT(hangAfterIndexBuildDumpsInsertsFromBulk)) { + if (MONGO_unlikely(hangAfterIndexBuildDumpsInsertsFromBulk.shouldFail())) { log() << "Hanging after dumping inserts from bulk builder"; - MONGO_FAIL_POINT_PAUSE_WHILE_SET(hangAfterIndexBuildDumpsInsertsFromBulk); + hangAfterIndexBuildDumpsInsertsFromBulk.pauseWhileSet(); } // Perform the first drain while holding an intent lock. @@ -908,9 +908,9 @@ void IndexBuildsCoordinator::_buildIndex( opCtx, replState->buildUUID, RecoveryUnit::ReadSource::kUnset)); } - if (MONGO_FAIL_POINT(hangAfterIndexBuildFirstDrain)) { + if (MONGO_unlikely(hangAfterIndexBuildFirstDrain.shouldFail())) { log() << "Hanging after index build first drain"; - MONGO_FAIL_POINT_PAUSE_WHILE_SET(hangAfterIndexBuildFirstDrain); + hangAfterIndexBuildFirstDrain.pauseWhileSet(); } // Perform the second drain while stopping writes on the collection. @@ -922,9 +922,9 @@ void IndexBuildsCoordinator::_buildIndex( opCtx, replState->buildUUID, RecoveryUnit::ReadSource::kUnset)); } - if (MONGO_FAIL_POINT(hangAfterIndexBuildSecondDrain)) { + if (MONGO_unlikely(hangAfterIndexBuildSecondDrain.shouldFail())) { log() << "Hanging after index build second drain"; - MONGO_FAIL_POINT_PAUSE_WHILE_SET(hangAfterIndexBuildSecondDrain); + hangAfterIndexBuildSecondDrain.pauseWhileSet(); } // Need to return the collection lock back to exclusive mode, to complete the index build. diff --git a/src/mongo/db/index_builds_coordinator.h b/src/mongo/db/index_builds_coordinator.h index e75292f004f..ff2f75fd288 100644 --- a/src/mongo/db/index_builds_coordinator.h +++ b/src/mongo/db/index_builds_coordinator.h @@ -507,8 +507,8 @@ private: // These fail points are used to control index build progress. Declared here to be shared // temporarily between createIndexes command and IndexBuildsCoordinator. -MONGO_FAIL_POINT_DECLARE(hangAfterIndexBuildFirstDrain); -MONGO_FAIL_POINT_DECLARE(hangAfterIndexBuildSecondDrain); -MONGO_FAIL_POINT_DECLARE(hangAfterIndexBuildDumpsInsertsFromBulk); +extern FailPoint hangAfterIndexBuildFirstDrain; +extern FailPoint hangAfterIndexBuildSecondDrain; +extern FailPoint hangAfterIndexBuildDumpsInsertsFromBulk; } // namespace mongo diff --git a/src/mongo/db/index_builds_coordinator_mongod.cpp b/src/mongo/db/index_builds_coordinator_mongod.cpp index 695edb86852..31a9859b38c 100644 --- a/src/mongo/db/index_builds_coordinator_mongod.cpp +++ b/src/mongo/db/index_builds_coordinator_mongod.cpp @@ -184,7 +184,7 @@ IndexBuildsCoordinatorMongod::startIndexBuild(OperationContext* opCtx, return; } - MONGO_FAIL_POINT_PAUSE_WHILE_SET(hangAfterInitializingIndexBuild); + hangAfterInitializingIndexBuild.pauseWhileSet(); auto opCtx = Client::getCurrent()->makeOperationContext(); diff --git a/src/mongo/db/key_generator.cpp b/src/mongo/db/key_generator.cpp index ab4eb8ef11e..def0b6c4ac5 100644 --- a/src/mongo/db/key_generator.cpp +++ b/src/mongo/db/key_generator.cpp @@ -76,7 +76,7 @@ KeyGenerator::KeyGenerator(std::string purpose, Status KeyGenerator::generateNewKeysIfNeeded(OperationContext* opCtx) { - if (MONGO_FAIL_POINT(disableKeyGeneration)) { + if (MONGO_unlikely(disableKeyGeneration.shouldFail())) { return {ErrorCodes::FailPointEnabled, "key generation disabled"}; } diff --git a/src/mongo/db/keys_collection_manager.cpp b/src/mongo/db/keys_collection_manager.cpp index a771a9150ad..bbad4f450c6 100644 --- a/src/mongo/db/keys_collection_manager.cpp +++ b/src/mongo/db/keys_collection_manager.cpp @@ -265,13 +265,9 @@ void KeysCollectionManager::PeriodicRunner::_doPeriodicRefresh(ServiceContext* s } } - MONGO_FAIL_POINT_BLOCK(maxKeyRefreshWaitTimeOverrideMS, data) { - const BSONObj& dataObj = data.getData(); - auto overrideMS = Milliseconds(dataObj["overrideMS"].numberInt()); - if (nextWakeup > overrideMS) { - nextWakeup = overrideMS; - } - } + maxKeyRefreshWaitTimeOverrideMS.execute([&](const BSONObj& data) { + nextWakeup = std::min(nextWakeup, Milliseconds(data["overrideMS"].numberInt())); + }); stdx::unique_lock<stdx::mutex> lock(_mutex); diff --git a/src/mongo/db/matcher/expression.h b/src/mongo/db/matcher/expression.h index f5bae2f7c55..b476db91330 100644 --- a/src/mongo/db/matcher/expression.h +++ b/src/mongo/db/matcher/expression.h @@ -46,7 +46,7 @@ namespace mongo { * Enabling the disableMatchExpressionOptimization fail point will stop match expressions from * being optimized. */ -MONGO_FAIL_POINT_DECLARE(disableMatchExpressionOptimization); +extern FailPoint disableMatchExpressionOptimization; class CollatorInterface; class MatchExpression; @@ -143,7 +143,7 @@ public: static std::unique_ptr<MatchExpression> optimize(std::unique_ptr<MatchExpression> expression) { // If the disableMatchExpressionOptimization failpoint is enabled, optimizations are skipped // and the expression is left unmodified. - if (MONGO_FAIL_POINT(disableMatchExpressionOptimization)) { + if (MONGO_unlikely(disableMatchExpressionOptimization.shouldFail())) { return expression; } diff --git a/src/mongo/db/matcher/expression_expr.cpp b/src/mongo/db/matcher/expression_expr.cpp index f356c4aa25b..62a345ab515 100644 --- a/src/mongo/db/matcher/expression_expr.cpp +++ b/src/mongo/db/matcher/expression_expr.cpp @@ -62,7 +62,7 @@ bool ExprMatchExpression::matches(const MatchableDocument* doc, MatchDetails* de auto value = _expression->evaluate(document, &variables); return value.coerceToBool(); } catch (const DBException&) { - if (MONGO_FAIL_POINT(ExprMatchExpressionMatchesReturnsFalseOnException)) { + if (MONGO_unlikely(ExprMatchExpressionMatchesReturnsFalseOnException.shouldFail())) { return false; } diff --git a/src/mongo/db/op_observer_impl.cpp b/src/mongo/db/op_observer_impl.cpp index 4ba8d17eea1..2ccfa6fa5dc 100644 --- a/src/mongo/db/op_observer_impl.cpp +++ b/src/mongo/db/op_observer_impl.cpp @@ -413,16 +413,18 @@ void OpObserverImpl::onInserts(OperationContext* opCtx, } void OpObserverImpl::onUpdate(OperationContext* opCtx, const OplogUpdateEntryArgs& args) { - MONGO_FAIL_POINT_BLOCK(failCollectionUpdates, extraData) { - auto collElem = extraData.getData()["collectionNS"]; - // If the failpoint specifies no collection or matches the existing one, fail. - if (!collElem || args.nss.ns() == collElem.String()) { + failCollectionUpdates.executeIf( + [&](const BSONObj&) { uasserted(40654, str::stream() << "failCollectionUpdates failpoint enabled, namespace: " << args.nss.ns() << ", update: " << args.updateArgs.update << " on document with " << args.updateArgs.criteria); - } - } + }, + [&](const BSONObj& data) { + // If the failpoint specifies no collection or matches the existing one, fail. + auto collElem = data["collectionNS"]; + return !collElem || args.nss.ns() == collElem.String(); + }); // Do not log a no-op operation; see SERVER-21738 if (args.updateArgs.update.isEmpty()) { diff --git a/src/mongo/db/operation_context.cpp b/src/mongo/db/operation_context.cpp index c3fef5535ae..6c21a7b58a1 100644 --- a/src/mongo/db/operation_context.cpp +++ b/src/mongo/db/operation_context.cpp @@ -139,10 +139,10 @@ bool OperationContext::hasDeadlineExpired() const { if (!hasDeadline()) { return false; } - if (MONGO_FAIL_POINT(maxTimeNeverTimeOut)) { + if (MONGO_unlikely(maxTimeNeverTimeOut.shouldFail())) { return false; } - if (MONGO_FAIL_POINT(maxTimeAlwaysTimeOut)) { + if (MONGO_unlikely(maxTimeAlwaysTimeOut.shouldFail())) { return true; } @@ -216,12 +216,12 @@ Status OperationContext::checkForInterruptNoAssert() noexcept { return Status::OK(); } - MONGO_FAIL_POINT_BLOCK(checkForInterruptFail, scopedFailPoint) { - if (opShouldFail(getClient(), scopedFailPoint.getData())) { + checkForInterruptFail.executeIf( + [&](auto&&) { log() << "set pending kill on op " << getOpID() << ", for checkForInterruptFail"; markKilled(); - } - } + }, + [&](auto&& data) { return opShouldFail(getClient(), data); }); const auto killStatus = getKillStatus(); if (killStatus != ErrorCodes::OK) { @@ -281,7 +281,7 @@ StatusWith<stdx::cv_status> OperationContext::waitForConditionOrInterruptNoAsser // maxTimeNeverTimeOut is set) then we assume that the incongruity is due to a clock mismatch // and return _timeoutError regardless. To prevent this behaviour, only consider the op's // deadline in the event that the maxTimeNeverTimeOut failpoint is not set. - bool opHasDeadline = (hasDeadline() && !MONGO_FAIL_POINT(maxTimeNeverTimeOut)); + bool opHasDeadline = (hasDeadline() && !MONGO_unlikely(maxTimeNeverTimeOut.shouldFail())); if (opHasDeadline) { deadline = std::min(deadline, getDeadline()); diff --git a/src/mongo/db/ops/write_ops_exec.cpp b/src/mongo/db/ops/write_ops_exec.cpp index 3b87ce8038d..8e46b008110 100644 --- a/src/mongo/db/ops/write_ops_exec.cpp +++ b/src/mongo/db/ops/write_ops_exec.cpp @@ -368,7 +368,7 @@ bool insertBatchAndHandleErrors(OperationContext* opCtx, true, // Check for interrupt periodically. wholeOp.getNamespace()); - if (MONGO_FAIL_POINT(failAllInserts)) { + if (MONGO_unlikely(failAllInserts.shouldFail())) { uasserted(ErrorCodes::InternalError, "failAllInserts failpoint active!"); } @@ -601,7 +601,7 @@ static SingleWriteResult performSingleUpdateOp(OperationContext* opCtx, false /*checkForInterrupt*/, ns); - if (MONGO_FAIL_POINT(failAllUpdates)) { + if (MONGO_unlikely(failAllUpdates.shouldFail())) { uasserted(ErrorCodes::InternalError, "failAllUpdates failpoint active!"); } @@ -848,7 +848,7 @@ static SingleWriteResult performSingleDeleteOp(OperationContext* opCtx, }, true // Check for interrupt periodically. ); - if (MONGO_FAIL_POINT(failAllRemoves)) { + if (MONGO_unlikely(failAllRemoves.shouldFail())) { uasserted(ErrorCodes::InternalError, "failAllRemoves failpoint active!"); } @@ -936,12 +936,12 @@ WriteResult performDeletes(OperationContext* opCtx, const write_ops::Delete& who curOp.setCommand_inlock(cmd); } ON_BLOCK_EXIT([&] { - if (MONGO_FAIL_POINT(hangBeforeChildRemoveOpFinishes)) { + if (MONGO_unlikely(hangBeforeChildRemoveOpFinishes.shouldFail())) { CurOpFailpointHelpers::waitWhileFailPointEnabled( &hangBeforeChildRemoveOpFinishes, opCtx, "hangBeforeChildRemoveOpFinishes"); } finishCurOp(opCtx, &curOp); - if (MONGO_FAIL_POINT(hangBeforeChildRemoveOpIsPopped)) { + if (MONGO_unlikely(hangBeforeChildRemoveOpIsPopped.shouldFail())) { CurOpFailpointHelpers::waitWhileFailPointEnabled( &hangBeforeChildRemoveOpIsPopped, opCtx, "hangBeforeChildRemoveOpIsPopped"); } @@ -959,7 +959,7 @@ WriteResult performDeletes(OperationContext* opCtx, const write_ops::Delete& who } } - if (MONGO_FAIL_POINT(hangAfterAllChildRemoveOpsArePopped)) { + if (MONGO_unlikely(hangAfterAllChildRemoveOpsArePopped.shouldFail())) { CurOpFailpointHelpers::waitWhileFailPointEnabled( &hangAfterAllChildRemoveOpsArePopped, opCtx, "hangAfterAllChildRemoveOpsArePopped"); } diff --git a/src/mongo/db/pipeline/document_source_cursor.cpp b/src/mongo/db/pipeline/document_source_cursor.cpp index 277db77429c..669cf02f670 100644 --- a/src/mongo/db/pipeline/document_source_cursor.cpp +++ b/src/mongo/db/pipeline/document_source_cursor.cpp @@ -86,7 +86,7 @@ void DocumentSourceCursor::loadBatch() { return; } - while (MONGO_FAIL_POINT(hangBeforeDocumentSourceCursorLoadBatch)) { + while (MONGO_unlikely(hangBeforeDocumentSourceCursorLoadBatch.shouldFail())) { log() << "Hanging aggregation due to 'hangBeforeDocumentSourceCursorLoadBatch' failpoint"; sleepmillis(10); } diff --git a/src/mongo/db/pipeline/document_source_exchange.cpp b/src/mongo/db/pipeline/document_source_exchange.cpp index 2220a73b496..76ca6839e96 100644 --- a/src/mongo/db/pipeline/document_source_exchange.cpp +++ b/src/mongo/db/pipeline/document_source_exchange.cpp @@ -315,7 +315,7 @@ DocumentSource::GetNextResult Exchange::getNext(OperationContext* opCtx, // The return value is an index of a full consumer buffer. size_t fullConsumerId = loadNextBatch(); - if (MONGO_FAIL_POINT(exchangeFailLoadNextBatch)) { + if (MONGO_unlikely(exchangeFailLoadNextBatch.shouldFail())) { log() << "exchangeFailLoadNextBatch fail point enabled."; uasserted(ErrorCodes::FailPointEnabled, "Asserting on loading the next batch due to failpoint."); diff --git a/src/mongo/db/pipeline/pipeline.cpp b/src/mongo/db/pipeline/pipeline.cpp index 355bc8a429a..c5fdbc032fb 100644 --- a/src/mongo/db/pipeline/pipeline.cpp +++ b/src/mongo/db/pipeline/pipeline.cpp @@ -266,7 +266,7 @@ void Pipeline::validateCommon() const { void Pipeline::optimizePipeline() { // If the disablePipelineOptimization failpoint is enabled, the pipeline won't be optimized. - if (MONGO_FAIL_POINT(disablePipelineOptimization)) { + if (MONGO_unlikely(disablePipelineOptimization.shouldFail())) { return; } diff --git a/src/mongo/db/pipeline/sharded_agg_helpers.cpp b/src/mongo/db/pipeline/sharded_agg_helpers.cpp index c5cec04fb34..3ef81811d71 100644 --- a/src/mongo/db/pipeline/sharded_agg_helpers.cpp +++ b/src/mongo/db/pipeline/sharded_agg_helpers.cpp @@ -425,10 +425,10 @@ std::vector<RemoteCursor> establishShardCursors( appendDbVersionIfPresent(cmdObjWithShardVersion, routingInfo->db())); } - if (MONGO_FAIL_POINT(clusterAggregateHangBeforeEstablishingShardCursors)) { + if (MONGO_unlikely(clusterAggregateHangBeforeEstablishingShardCursors.shouldFail())) { log() << "clusterAggregateHangBeforeEstablishingShardCursors fail point enabled. Blocking " "until fail point is disabled."; - while (MONGO_FAIL_POINT(clusterAggregateHangBeforeEstablishingShardCursors)) { + while (MONGO_unlikely(clusterAggregateHangBeforeEstablishingShardCursors.shouldFail())) { sleepsecs(1); } } diff --git a/src/mongo/db/query/find.cpp b/src/mongo/db/query/find.cpp index 65816afeef4..2f351c930d3 100644 --- a/src/mongo/db/query/find.cpp +++ b/src/mongo/db/query/find.cpp @@ -248,7 +248,7 @@ Message getMore(OperationContext* opCtx, curOp.ensureStarted(); // For testing, we may want to fail if we receive a getmore. - if (MONGO_FAIL_POINT(failReceivedGetmore)) { + if (MONGO_unlikely(failReceivedGetmore.shouldFail())) { MONGO_UNREACHABLE; } @@ -369,8 +369,7 @@ Message getMore(OperationContext* opCtx, // repeatedly release and re-acquire the collection readLock at regular intervals until // the failpoint is released. This is done in order to avoid deadlocks caused by the // pinned-cursor failpoints in this file (see SERVER-21997). - MONGO_FAIL_POINT_BLOCK(waitAfterPinningCursorBeforeGetMoreBatch, options) { - const BSONObj& data = options.getData(); + waitAfterPinningCursorBeforeGetMoreBatch.execute([&](const BSONObj& data) { if (data["shouldNotdropLock"].booleanSafe()) { dropAndReaquireReadLock = []() {}; } @@ -381,7 +380,7 @@ Message getMore(OperationContext* opCtx, dropAndReaquireReadLock, false, nss); - } + }); const auto replicationMode = repl::ReplicationCoordinator::get(opCtx)->getReplicationMode(); @@ -459,7 +458,7 @@ Message getMore(OperationContext* opCtx, // accumulate over the course of a cursor's lifetime. PlanSummaryStats preExecutionStats; Explain::getSummaryStats(*exec, &preExecutionStats); - if (MONGO_FAIL_POINT(waitWithPinnedCursorDuringGetMoreBatch)) { + if (MONGO_unlikely(waitWithPinnedCursorDuringGetMoreBatch.shouldFail())) { CurOpFailpointHelpers::waitWhileFailPointEnabled(&waitWithPinnedCursorDuringGetMoreBatch, opCtx, "waitWithPinnedCursorDuringGetMoreBatch", @@ -548,7 +547,7 @@ Message getMore(OperationContext* opCtx, // If the 'waitBeforeUnpinningOrDeletingCursorAfterGetMoreBatch' failpoint is active, we // set the 'msg' field of this operation's CurOp to signal that we've hit this point and // then spin until the failpoint is released. - if (MONGO_FAIL_POINT(waitBeforeUnpinningOrDeletingCursorAfterGetMoreBatch)) { + if (MONGO_unlikely(waitBeforeUnpinningOrDeletingCursorAfterGetMoreBatch.shouldFail())) { CurOpFailpointHelpers::waitWhileFailPointEnabled( &waitBeforeUnpinningOrDeletingCursorAfterGetMoreBatch, opCtx, diff --git a/src/mongo/db/query/find_common.h b/src/mongo/db/query/find_common.h index 14515224326..90216d9138c 100644 --- a/src/mongo/db/query/find_common.h +++ b/src/mongo/db/query/find_common.h @@ -57,23 +57,23 @@ class CanonicalQuery; class QueryRequest; // Failpoint for making find hang. -MONGO_FAIL_POINT_DECLARE(waitInFindBeforeMakingBatch); +extern FailPoint waitInFindBeforeMakingBatch; // Failpoint for making getMore not wait for an awaitdata cursor. Allows us to avoid waiting during // tests. -MONGO_FAIL_POINT_DECLARE(disableAwaitDataForGetMoreCmd); +extern FailPoint disableAwaitDataForGetMoreCmd; // Enabling this fail point will cause getMores to busy wait after pinning the cursor // but before we have started building the batch, until the fail point is disabled. -MONGO_FAIL_POINT_DECLARE(waitAfterPinningCursorBeforeGetMoreBatch); +extern FailPoint waitAfterPinningCursorBeforeGetMoreBatch; // Enabling this fail point will cause getMores to busy wait after setting up the plan executor and // before beginning the batch. -MONGO_FAIL_POINT_DECLARE(waitWithPinnedCursorDuringGetMoreBatch); +extern FailPoint waitWithPinnedCursorDuringGetMoreBatch; // Enabling this failpoint will cause getMores to wait just before it unpins its cursor after it // has completed building the current batch. -MONGO_FAIL_POINT_DECLARE(waitBeforeUnpinningOrDeletingCursorAfterGetMoreBatch); +extern FailPoint waitBeforeUnpinningOrDeletingCursorAfterGetMoreBatch; /** * Suite of find/getMore related functions used in both the mongod and mongos query paths. diff --git a/src/mongo/db/query/plan_executor_impl.cpp b/src/mongo/db/query/plan_executor_impl.cpp index b3f73cf26bc..7a07b1f2a03 100644 --- a/src/mongo/db/query/plan_executor_impl.cpp +++ b/src/mongo/db/query/plan_executor_impl.cpp @@ -472,7 +472,7 @@ PlanExecutor::ExecState PlanExecutorImpl::_waitForInserts(CappedInsertNotifierDa PlanExecutor::ExecState PlanExecutorImpl::_getNextImpl(Snapshotted<BSONObj>* objOut, RecordId* dlOut) { - if (MONGO_FAIL_POINT(planExecutorAlwaysFails)) { + if (MONGO_unlikely(planExecutorAlwaysFails.shouldFail())) { Status status(ErrorCodes::InternalError, str::stream() << "PlanExecutor hit planExecutorAlwaysFails fail point"); *objOut = @@ -573,7 +573,8 @@ PlanExecutor::ExecState PlanExecutorImpl::_getNextImpl(Snapshotted<BSONObj>* obj // This result didn't have the data the caller wanted, try again. } else if (PlanStage::NEED_YIELD == code) { invariant(id == WorkingSet::INVALID_ID); - if (!_yieldPolicy->canAutoYield() || MONGO_FAIL_POINT(skipWriteConflictRetries)) { + if (!_yieldPolicy->canAutoYield() || + MONGO_unlikely(skipWriteConflictRetries.shouldFail())) { throw WriteConflictException(); } @@ -589,10 +590,10 @@ PlanExecutor::ExecState PlanExecutorImpl::_getNextImpl(Snapshotted<BSONObj>* obj } else if (PlanStage::NEED_TIME == code) { // Fall through to yield check at end of large conditional. } else if (PlanStage::IS_EOF == code) { - if (MONGO_FAIL_POINT(planExecutorHangBeforeShouldWaitForInserts)) { + if (MONGO_unlikely(planExecutorHangBeforeShouldWaitForInserts.shouldFail())) { log() << "PlanExecutor - planExecutorHangBeforeShouldWaitForInserts fail point " "enabled. Blocking until fail point is disabled."; - MONGO_FAIL_POINT_PAUSE_WHILE_SET(planExecutorHangBeforeShouldWaitForInserts); + planExecutorHangBeforeShouldWaitForInserts.pauseWhileSet(); } if (!_shouldWaitForInserts()) { return PlanExecutor::IS_EOF; diff --git a/src/mongo/db/query/plan_yield_policy.cpp b/src/mongo/db/query/plan_yield_policy.cpp index 43590f27d7d..fbdb0d117ca 100644 --- a/src/mongo/db/query/plan_yield_policy.cpp +++ b/src/mongo/db/query/plan_yield_policy.cpp @@ -90,7 +90,7 @@ Status PlanYieldPolicy::yieldOrInterrupt(std::function<void()> whileYieldingFn) invariant(opCtx); // If the 'setInterruptOnlyPlansCheckForInterruptHang' fail point is enabled, set the 'msg' // field of this operation's CurOp to signal that we've hit this point. - if (MONGO_FAIL_POINT(setInterruptOnlyPlansCheckForInterruptHang)) { + if (MONGO_unlikely(setInterruptOnlyPlansCheckForInterruptHang.shouldFail())) { CurOpFailpointHelpers::waitWhileFailPointEnabled( &setInterruptOnlyPlansCheckForInterruptHang, opCtx, @@ -183,20 +183,17 @@ void PlanYieldPolicy::_yieldAllLocks(OperationContext* opCtx, // Track the number of yields in CurOp. CurOp::get(opCtx)->yielded(); - MONGO_FAIL_POINT_BLOCK(setYieldAllLocksHang, config) { - StringData ns{config.getData().getStringField("namespace")}; - if (ns.empty() || ns == planExecNS.ns()) { - MONGO_FAIL_POINT_PAUSE_WHILE_SET(setYieldAllLocksHang); - } - } - - MONGO_FAIL_POINT_BLOCK(setYieldAllLocksWait, customWait) { - const BSONObj& data = customWait.getData(); - BSONElement customWaitNS = data["namespace"]; - if (!customWaitNS || planExecNS.ns() == customWaitNS.str()) { - sleepFor(Milliseconds(data["waitForMillis"].numberInt())); - } - } + setYieldAllLocksHang.executeIf([](auto&&) { setYieldAllLocksHang.pauseWhileSet(); }, + [&](const BSONObj& config) { + StringData ns = config.getStringField("namespace"); + return ns.empty() || ns == planExecNS.ns(); + }); + setYieldAllLocksWait.executeIf( + [&](const BSONObj& data) { sleepFor(Milliseconds(data["waitForMillis"].numberInt())); }, + [&](const BSONObj& config) { + BSONElement dataNs = config["namespace"]; + return !dataNs || planExecNS.ns() == dataNs.str(); + }); if (whileYieldingFn) { whileYieldingFn(); diff --git a/src/mongo/db/repair_database_and_check_version.cpp b/src/mongo/db/repair_database_and_check_version.cpp index 1a0711faddf..b08d67696c0 100644 --- a/src/mongo/db/repair_database_and_check_version.cpp +++ b/src/mongo/db/repair_database_and_check_version.cpp @@ -363,7 +363,7 @@ bool repairDatabasesAndCheckVersion(OperationContext* opCtx) { if (storageGlobalParams.repair) { invariant(!storageGlobalParams.readOnly); - if (MONGO_FAIL_POINT(exitBeforeDataRepair)) { + if (MONGO_unlikely(exitBeforeDataRepair.shouldFail())) { log() << "Exiting because 'exitBeforeDataRepair' fail point was set."; quickExit(EXIT_ABRUPT); } @@ -420,7 +420,7 @@ bool repairDatabasesAndCheckVersion(OperationContext* opCtx) { } if (storageGlobalParams.repair) { - if (MONGO_FAIL_POINT(exitBeforeRepairInvalidatesConfig)) { + if (MONGO_unlikely(exitBeforeRepairInvalidatesConfig.shouldFail())) { log() << "Exiting because 'exitBeforeRepairInvalidatesConfig' fail point was set."; quickExit(EXIT_ABRUPT); } diff --git a/src/mongo/db/repl/apply_ops.cpp b/src/mongo/db/repl/apply_ops.cpp index 1e947868a65..6426f8e1979 100644 --- a/src/mongo/db/repl/apply_ops.cpp +++ b/src/mongo/db/repl/apply_ops.cpp @@ -271,8 +271,8 @@ Status _applyOps(OperationContext* opCtx, (*numApplied)++; - if (MONGO_FAIL_POINT(applyOpsPauseBetweenOperations)) { - MONGO_FAIL_POINT_PAUSE_WHILE_SET(applyOpsPauseBetweenOperations); + if (MONGO_unlikely(applyOpsPauseBetweenOperations.shouldFail())) { + applyOpsPauseBetweenOperations.pauseWhileSet(); } } diff --git a/src/mongo/db/repl/bgsync.cpp b/src/mongo/db/repl/bgsync.cpp index 990e7167aeb..870f9d54e45 100644 --- a/src/mongo/db/repl/bgsync.cpp +++ b/src/mongo/db/repl/bgsync.cpp @@ -219,7 +219,7 @@ void BackgroundSync::_runProducer() { } void BackgroundSync::_produce() { - if (MONGO_FAIL_POINT(stopReplProducer)) { + if (MONGO_unlikely(stopReplProducer.shouldFail())) { // This log output is used in js tests so please leave it. log() << "bgsync - stopReplProducer fail point " "enabled. Blocking until fail point is disabled."; @@ -227,7 +227,7 @@ void BackgroundSync::_produce() { // TODO(SERVER-27120): Remove the return statement and uncomment the while loop. // Currently we cannot block here or we prevent primaries from being fully elected since // we'll never call _signalNoNewDataForApplier. - // while (MONGO_FAIL_POINT(stopReplProducer) && !inShutdown()) { + // while (MONGO_unlikely(stopReplProducer.shouldFail()) && !inShutdown()) { // mongo::sleepsecs(1); // } mongo::sleepsecs(1); @@ -577,11 +577,11 @@ void BackgroundSync::_runRollback(OperationContext* opCtx, } } - if (MONGO_FAIL_POINT(rollbackHangBeforeStart)) { + if (MONGO_unlikely(rollbackHangBeforeStart.shouldFail())) { // This log output is used in js tests so please leave it. log() << "rollback - rollbackHangBeforeStart fail point " "enabled. Blocking until fail point is disabled."; - while (MONGO_FAIL_POINT(rollbackHangBeforeStart) && !inShutdown()) { + while (MONGO_unlikely(rollbackHangBeforeStart.shouldFail()) && !inShutdown()) { mongo::sleepsecs(1); } } diff --git a/src/mongo/db/repl/collection_cloner.cpp b/src/mongo/db/repl/collection_cloner.cpp index b081be1fce0..c5270c5370d 100644 --- a/src/mongo/db/repl/collection_cloner.cpp +++ b/src/mongo/db/repl/collection_cloner.cpp @@ -429,20 +429,22 @@ void CollectionCloner::_beginCollectionCallback(const executor::TaskExecutor::Ca _finishCallback(cbd.status); return; } - MONGO_FAIL_POINT_BLOCK(initialSyncHangCollectionClonerBeforeEstablishingCursor, nssData) { - const BSONObj& data = nssData.getData(); - auto nss = data["nss"].str(); - // Only hang when cloning the specified collection, or if no collection was specified. - if (nss.empty() || _destNss.toString() == nss) { - while (MONGO_FAIL_POINT(initialSyncHangCollectionClonerBeforeEstablishingCursor) && + initialSyncHangCollectionClonerBeforeEstablishingCursor.executeIf( + [&](const BSONObj& data) { + while (MONGO_unlikely( + initialSyncHangCollectionClonerBeforeEstablishingCursor.shouldFail()) && !_isShuttingDown()) { log() << "initialSyncHangCollectionClonerBeforeEstablishingCursor fail point " "enabled for " << _destNss.toString() << ". Blocking until fail point is disabled."; mongo::sleepsecs(1); } - } - } + }, + [&](const BSONObj& data) { + auto nss = data["nss"].str(); + // Only hang when cloning the specified collection, or if no collection was specified. + return nss.empty() || nss == _destNss.toString(); + }); if (!_idIndexSpec.isEmpty() && _options.autoIndexId == CollectionOptions::NO) { warning() << "Found the _id_ index spec but the collection specified autoIndexId of false on ns:" @@ -510,16 +512,16 @@ void CollectionCloner::_runQuery(const executor::TaskExecutor::CallbackArgs& cal } } - MONGO_FAIL_POINT_BLOCK(initialSyncHangBeforeCollectionClone, options) { - const BSONObj& data = options.getData(); - if (data["namespace"].String() == _destNss.ns()) { + initialSyncHangBeforeCollectionClone.executeIf( + [&](const BSONObj&) { log() << "initial sync - initialSyncHangBeforeCollectionClone fail point " "enabled. Blocking until fail point is disabled."; - while (MONGO_FAIL_POINT(initialSyncHangBeforeCollectionClone) && !_isShuttingDown()) { + while (MONGO_unlikely(initialSyncHangBeforeCollectionClone.shouldFail()) && + !_isShuttingDown()) { mongo::sleepsecs(1); } - } - } + }, + [&](const BSONObj& data) { return data["namespace"].String() == _destNss.ns(); }); Status clientConnectionStatus = _clientConnection->connect(_source, StringData()); if (!clientConnectionStatus.isOK()) { @@ -600,20 +602,22 @@ void CollectionCloner::_handleNextBatch(std::shared_ptr<OnCompletionGuard> onCom uassertStatusOK(newStatus); } - MONGO_FAIL_POINT_BLOCK(initialSyncHangCollectionClonerAfterHandlingBatchResponse, nssData) { - const BSONObj& data = nssData.getData(); - auto nss = data["nss"].str(); - // Only hang when cloning the specified collection, or if no collection was specified. - if (nss.empty() || _destNss.toString() == nss) { - while (MONGO_FAIL_POINT(initialSyncHangCollectionClonerAfterHandlingBatchResponse) && + initialSyncHangCollectionClonerAfterHandlingBatchResponse.executeIf( + [&](const BSONObj&) { + while (MONGO_unlikely( + initialSyncHangCollectionClonerAfterHandlingBatchResponse.shouldFail()) && !_isShuttingDown()) { log() << "initialSyncHangCollectionClonerAfterHandlingBatchResponse fail point " "enabled for " << _destNss.toString() << ". Blocking until fail point is disabled."; mongo::sleepsecs(1); } - } - } + }, + [&](const BSONObj& data) { + // Only hang when cloning the specified collection, or if no collection was specified. + auto nss = data["nss"].str(); + return nss.empty() || nss == _destNss.toString(); + }); } void CollectionCloner::_verifyCollectionWasDropped( @@ -702,19 +706,21 @@ void CollectionCloner::_insertDocumentsCallback( return; } - MONGO_FAIL_POINT_BLOCK(initialSyncHangDuringCollectionClone, options) { - const BSONObj& data = options.getData(); - if (data["namespace"].String() == _destNss.ns() && - static_cast<int>(_stats.documentsCopied) >= data["numDocsToClone"].numberInt()) { + initialSyncHangDuringCollectionClone.executeIf( + [&](const BSONObj&) { lk.unlock(); log() << "initial sync - initialSyncHangDuringCollectionClone fail point " "enabled. Blocking until fail point is disabled."; - while (MONGO_FAIL_POINT(initialSyncHangDuringCollectionClone) && !_isShuttingDown()) { + while (MONGO_unlikely(initialSyncHangDuringCollectionClone.shouldFail()) && + !_isShuttingDown()) { mongo::sleepsecs(1); } lk.lock(); - } - } + }, + [&](const BSONObj& data) { + return data["namespace"].String() == _destNss.ns() && + static_cast<int>(_stats.documentsCopied) >= data["numDocsToClone"].numberInt(); + }); } void CollectionCloner::_finishCallback(const Status& status) { diff --git a/src/mongo/db/repl/database_cloner.cpp b/src/mongo/db/repl/database_cloner.cpp index 5a26e80195d..63a00583854 100644 --- a/src/mongo/db/repl/database_cloner.cpp +++ b/src/mongo/db/repl/database_cloner.cpp @@ -178,19 +178,21 @@ Status DatabaseCloner::startup() noexcept { return Status(ErrorCodes::ShutdownInProgress, "database cloner completed"); } - MONGO_FAIL_POINT_BLOCK(initialSyncHangBeforeListCollections, customArgs) { - const auto& data = customArgs.getData(); - const auto databaseElem = data["database"]; - if (!databaseElem || databaseElem.checkAndGetStringData() == _dbname) { + initialSyncHangBeforeListCollections.executeIf( + [&](const BSONObj&) { lk.unlock(); log() << "initial sync - initialSyncHangBeforeListCollections fail point " "enabled. Blocking until fail point is disabled."; - while (MONGO_FAIL_POINT(initialSyncHangBeforeListCollections) && !_isShuttingDown()) { + while (MONGO_unlikely(initialSyncHangBeforeListCollections.shouldFail()) && + !_isShuttingDown()) { mongo::sleepsecs(1); } lk.lock(); - } - } + }, + [&](const BSONObj& data) { + const auto databaseElem = data["database"]; + return !databaseElem || databaseElem.checkAndGetStringData() == _dbname; + }); _stats.start = _executor->now(); LOG(1) << "Scheduling listCollections call for database: " << _dbname; @@ -294,16 +296,13 @@ void DatabaseCloner::_listCollectionsCallback(const StatusWith<Fetcher::QueryRes return; } - MONGO_FAIL_POINT_BLOCK(initialSyncHangAfterListCollections, options) { - const BSONObj& data = options.getData(); - if (data["database"].String() == _dbname) { + initialSyncHangAfterListCollections.executeIf( + [&](const BSONObj&) { log() << "initial sync - initialSyncHangAfterListCollections fail point " "enabled. Blocking until fail point is disabled."; - while (MONGO_FAIL_POINT(initialSyncHangAfterListCollections)) { - mongo::sleepsecs(1); - } - } - } + initialSyncHangAfterListCollections.pauseWhileSet(); + }, + [&](const BSONObj& data) { return data["database"].String() == _dbname; }); _collectionNamespaces.reserve(_collectionInfos.size()); std::set<std::string> seen; diff --git a/src/mongo/db/repl/initial_syncer.cpp b/src/mongo/db/repl/initial_syncer.cpp index 418899be19c..444435d4085 100644 --- a/src/mongo/db/repl/initial_syncer.cpp +++ b/src/mongo/db/repl/initial_syncer.cpp @@ -542,7 +542,7 @@ void InitialSyncer::_chooseSyncSourceCallback( return; } - if (MONGO_FAIL_POINT(failInitialSyncWithBadHost)) { + if (MONGO_unlikely(failInitialSyncWithBadHost.shouldFail())) { status = Status(ErrorCodes::InvalidSyncSource, "initial sync failed - failInitialSyncWithBadHost failpoint is set."); onCompletionGuard->setResultAndCancelRemainingWork_inlock(lock, status); @@ -580,12 +580,13 @@ void InitialSyncer::_chooseSyncSourceCallback( return; } - if (MONGO_FAIL_POINT(initialSyncHangBeforeCreatingOplog)) { + if (MONGO_unlikely(initialSyncHangBeforeCreatingOplog.shouldFail())) { // This log output is used in js tests so please leave it. log() << "initial sync - initialSyncHangBeforeCreatingOplog fail point " "enabled. Blocking until fail point is disabled."; lock.unlock(); - while (MONGO_FAIL_POINT(initialSyncHangBeforeCreatingOplog) && !_isShuttingDown()) { + while (MONGO_unlikely(initialSyncHangBeforeCreatingOplog.shouldFail()) && + !_isShuttingDown()) { mongo::sleepsecs(1); } lock.lock(); @@ -953,14 +954,15 @@ void InitialSyncer::_fcvFetcherCallback(const StatusWith<Fetcher::QueryResponse> return; } - if (MONGO_FAIL_POINT(initialSyncHangBeforeCopyingDatabases)) { + if (MONGO_unlikely(initialSyncHangBeforeCopyingDatabases.shouldFail())) { lock.unlock(); // This could have been done with a scheduleWorkAt but this is used only by JS tests where // we run with multiple threads so it's fine to spin on this thread. // This log output is used in js tests so please leave it. log() << "initial sync - initialSyncHangBeforeCopyingDatabases fail point " "enabled. Blocking until fail point is disabled."; - while (MONGO_FAIL_POINT(initialSyncHangBeforeCopyingDatabases) && !_isShuttingDown()) { + while (MONGO_unlikely(initialSyncHangBeforeCopyingDatabases.shouldFail()) && + !_isShuttingDown()) { mongo::sleepsecs(1); } lock.lock(); @@ -1026,13 +1028,13 @@ void InitialSyncer::_databasesClonerCallback(const Status& databaseClonerFinishS log() << "Finished cloning data: " << redact(databaseClonerFinishStatus) << ". Beginning oplog replay."; - if (MONGO_FAIL_POINT(initialSyncHangAfterDataCloning)) { + if (MONGO_unlikely(initialSyncHangAfterDataCloning.shouldFail())) { // This could have been done with a scheduleWorkAt but this is used only by JS tests where // we run with multiple threads so it's fine to spin on this thread. // This log output is used in js tests so please leave it. log() << "initial sync - initialSyncHangAfterDataCloning fail point " "enabled. Blocking until fail point is disabled."; - while (MONGO_FAIL_POINT(initialSyncHangAfterDataCloning) && !_isShuttingDown()) { + while (MONGO_unlikely(initialSyncHangAfterDataCloning.shouldFail()) && !_isShuttingDown()) { mongo::sleepsecs(1); } } @@ -1147,22 +1149,22 @@ void InitialSyncer::_getNextApplierBatchCallback( // Set and unset by the InitialSyncTest fixture to cause initial sync to pause so that the // Initial Sync Fuzzer can run commands on the sync source. - if (MONGO_FAIL_POINT(initialSyncFuzzerSynchronizationPoint1)) { + if (MONGO_unlikely(initialSyncFuzzerSynchronizationPoint1.shouldFail())) { log() << "Initial Syncer is about to apply the next oplog batch of size: " << batchResult.getValue().size(); log() << "initialSyncFuzzerSynchronizationPoint1 fail point enabled."; - MONGO_FAIL_POINT_PAUSE_WHILE_SET(initialSyncFuzzerSynchronizationPoint1); + initialSyncFuzzerSynchronizationPoint1.pauseWhileSet(); } - if (MONGO_FAIL_POINT(initialSyncFuzzerSynchronizationPoint2)) { + if (MONGO_unlikely(initialSyncFuzzerSynchronizationPoint2.shouldFail())) { log() << "initialSyncFuzzerSynchronizationPoint2 fail point enabled."; - MONGO_FAIL_POINT_PAUSE_WHILE_SET(initialSyncFuzzerSynchronizationPoint2); + initialSyncFuzzerSynchronizationPoint2.pauseWhileSet(); } - if (MONGO_FAIL_POINT(failInitialSyncBeforeApplyingBatch)) { + if (MONGO_unlikely(failInitialSyncBeforeApplyingBatch.shouldFail())) { log() << "initial sync - failInitialSyncBeforeApplyingBatch fail point enabled. Pausing" << "until fail point is disabled, then will fail initial sync."; - MONGO_FAIL_POINT_PAUSE_WHILE_SET(failInitialSyncBeforeApplyingBatch); + failInitialSyncBeforeApplyingBatch.pauseWhileSet(); status = Status(ErrorCodes::CallbackCanceled, "failInitialSyncBeforeApplyingBatch fail point enabled"); onCompletionGuard->setResultAndCancelRemainingWork_inlock(lock, status); @@ -1236,7 +1238,7 @@ void InitialSyncer::_multiApplierCallback(const Status& multiApplierStatus, // Set to cause initial sync to fassert instead of restart if applying a batch fails, so that // tests can be robust to network errors but not oplog idempotency errors. - if (MONGO_FAIL_POINT(initialSyncFassertIfApplyingBatchFails)) { + if (MONGO_unlikely(initialSyncFassertIfApplyingBatchFails.shouldFail())) { log() << "initialSyncFassertIfApplyingBatchFails fail point enabled."; fassert(31210, status); } @@ -1325,9 +1327,9 @@ void InitialSyncer::_finishInitialSyncAttempt(const StatusWith<OpTimeAndWallTime _stats.initialSyncAttemptInfos.emplace_back( InitialSyncer::InitialSyncAttemptInfo{runTime, result.getStatus(), _syncSource}); - if (MONGO_FAIL_POINT(failAndHangInitialSync)) { + if (MONGO_unlikely(failAndHangInitialSync.shouldFail())) { log() << "failAndHangInitialSync fail point enabled."; - MONGO_FAIL_POINT_PAUSE_WHILE_SET(failAndHangInitialSync); + failAndHangInitialSync.pauseWhileSet(); result = Status(ErrorCodes::InternalError, "failAndHangInitialSync fail point enabled"); } @@ -1399,11 +1401,11 @@ void InitialSyncer::_finishCallback(StatusWith<OpTimeAndWallTime> lastApplied) { std::swap(_onCompletion, onCompletion); } - if (MONGO_FAIL_POINT(initialSyncHangBeforeFinish)) { + if (MONGO_unlikely(initialSyncHangBeforeFinish.shouldFail())) { // This log output is used in js tests so please leave it. log() << "initial sync - initialSyncHangBeforeFinish fail point " "enabled. Blocking until fail point is disabled."; - while (MONGO_FAIL_POINT(initialSyncHangBeforeFinish) && !_isShuttingDown()) { + while (MONGO_unlikely(initialSyncHangBeforeFinish.shouldFail()) && !_isShuttingDown()) { mongo::sleepsecs(1); } } @@ -1428,7 +1430,7 @@ void InitialSyncer::_finishCallback(StatusWith<OpTimeAndWallTime> lastApplied) { // Clear the initial sync progress after an initial sync attempt has been successfully // completed. - if (lastApplied.isOK() && !MONGO_FAIL_POINT(skipClearInitialSyncState)) { + if (lastApplied.isOK() && !MONGO_unlikely(skipClearInitialSyncState.shouldFail())) { _initialSyncState.reset(); } } @@ -1628,7 +1630,7 @@ StatusWith<Operations> InitialSyncer::_getNextApplierBatch_inlock() { // If the fail-point is active, delay the apply batch by returning an empty batch so that // _getNextApplierBatchCallback() will reschedule itself at a later time. // See InitialSyncerOptions::getApplierBatchCallbackRetryWait. - if (MONGO_FAIL_POINT(rsSyncApplyStop)) { + if (MONGO_unlikely(rsSyncApplyStop.shouldFail())) { return Operations(); } diff --git a/src/mongo/db/repl/initial_syncer.h b/src/mongo/db/repl/initial_syncer.h index 414de99a576..6ad23526d85 100644 --- a/src/mongo/db/repl/initial_syncer.h +++ b/src/mongo/db/repl/initial_syncer.h @@ -63,13 +63,13 @@ namespace repl { // TODO: Remove forward declares once we remove rs_initialsync.cpp and other dependents. // Failpoint which fails initial sync and leaves an oplog entry in the buffer. -MONGO_FAIL_POINT_DECLARE(failInitSyncWithBufferedEntriesLeft); +extern FailPoint failInitSyncWithBufferedEntriesLeft; // Failpoint which causes the initial sync function to hang before copying databases. -MONGO_FAIL_POINT_DECLARE(initialSyncHangBeforeCopyingDatabases); +extern FailPoint initialSyncHangBeforeCopyingDatabases; // Failpoint which stops the applier. -MONGO_FAIL_POINT_DECLARE(rsSyncApplyStop); +extern FailPoint rsSyncApplyStop; struct InitialSyncState; struct MemberState; diff --git a/src/mongo/db/repl/initial_syncer_test.cpp b/src/mongo/db/repl/initial_syncer_test.cpp index 55f4b1074bf..9d95ee96821 100644 --- a/src/mongo/db/repl/initial_syncer_test.cpp +++ b/src/mongo/db/repl/initial_syncer_test.cpp @@ -3495,7 +3495,7 @@ TEST_F( // Enable 'rsSyncApplyStop' so that _getNextApplierBatch_inlock() returns an empty batch of // operations instead of a batch containing an oplog entry with a bad version. - auto failPoint = getGlobalFailPointRegistry()->getFailPoint("rsSyncApplyStop"); + auto failPoint = globalFailPointRegistry().find("rsSyncApplyStop"); failPoint->setMode(FailPoint::alwaysOn); ON_BLOCK_EXIT([failPoint]() { failPoint->setMode(FailPoint::off); }); @@ -3954,7 +3954,7 @@ TEST_F(InitialSyncerTest, auto opCtx = makeOpCtx(); // This fail point makes chooseSyncSourceCallback fail with an InvalidSyncSource error. - auto failPoint = getGlobalFailPointRegistry()->getFailPoint("failInitialSyncWithBadHost"); + auto failPoint = globalFailPointRegistry().find("failInitialSyncWithBadHost"); failPoint->setMode(FailPoint::alwaysOn); ON_BLOCK_EXIT([failPoint]() { failPoint->setMode(FailPoint::off); }); diff --git a/src/mongo/db/repl/oplog.cpp b/src/mongo/db/repl/oplog.cpp index 988076d25f0..fa388531db3 100644 --- a/src/mongo/db/repl/oplog.cpp +++ b/src/mongo/db/repl/oplog.cpp @@ -339,10 +339,9 @@ void _logOpsInner(OperationContext* opCtx, } // Optionally hang before advancing lastApplied. - if (MONGO_FAIL_POINT(hangBeforeLogOpAdvancesLastApplied)) { + if (MONGO_unlikely(hangBeforeLogOpAdvancesLastApplied.shouldFail())) { log() << "hangBeforeLogOpAdvancesLastApplied fail point enabled."; - MONGO_FAIL_POINT_PAUSE_WHILE_SET_OR_INTERRUPTED(opCtx, - hangBeforeLogOpAdvancesLastApplied); + hangBeforeLogOpAdvancesLastApplied.pauseWhileSet(opCtx); } // Optimes on the primary should always represent consistent database states. @@ -477,13 +476,12 @@ std::vector<OpTime> logInsertOps(OperationContext* opCtx, RecordId(), RecordData(bsonOplogEntries[i].objdata(), bsonOplogEntries[i].objsize())}; } - MONGO_FAIL_POINT_BLOCK(sleepBetweenInsertOpTimeGenerationAndLogOp, customWait) { - const BSONObj& data = customWait.getData(); + sleepBetweenInsertOpTimeGenerationAndLogOp.execute([&](const BSONObj& data) { auto numMillis = data["waitForMillis"].numberInt(); log() << "Sleeping for " << numMillis << "ms after receiving " << count << " optimes from " << opTimes.front() << " to " << opTimes.back(); sleepmillis(numMillis); - } + }); invariant(!opTimes.empty()); auto lastOpTime = opTimes.back(); diff --git a/src/mongo/db/repl/oplog_fetcher.cpp b/src/mongo/db/repl/oplog_fetcher.cpp index 0aa50c4be16..89ada66c851 100644 --- a/src/mongo/db/repl/oplog_fetcher.cpp +++ b/src/mongo/db/repl/oplog_fetcher.cpp @@ -370,7 +370,7 @@ Milliseconds OplogFetcher::getAwaitDataTimeout_forTest() const { } Milliseconds OplogFetcher::_getGetMoreMaxTime() const { - if (MONGO_FAIL_POINT(setSmallOplogGetMoreMaxTimeMS)) { + if (MONGO_unlikely(setSmallOplogGetMoreMaxTimeMS.shouldFail())) { return Milliseconds(50); } @@ -382,24 +382,31 @@ StatusWith<BSONObj> OplogFetcher::_onSuccessfulBatch(const Fetcher::QueryRespons // Stop fetching and return on fail point. // This fail point makes the oplog fetcher ignore the downloaded batch of operations and not // error out. The FailPointEnabled error will be caught by the AbstractOplogFetcher. - if (MONGO_FAIL_POINT(stopReplProducer)) { + if (MONGO_unlikely(stopReplProducer.shouldFail())) { return Status(ErrorCodes::FailPointEnabled, "stopReplProducer fail point is enabled"); } // Stop fetching and return when we reach a particular document. This failpoint should be used // with the setParameter bgSyncOplogFetcherBatchSize=1, so that documents are fetched one at a // time. - MONGO_FAIL_POINT_BLOCK(stopReplProducerOnDocument, fp) { - auto opCtx = cc().makeOperationContext(); - boost::intrusive_ptr<ExpressionContext> expCtx(new ExpressionContext(opCtx.get(), nullptr)); - auto query = fp.getData()["document"].Obj(); - Matcher m(query, expCtx); - if (!queryResponse.documents.empty() && - m.matches(queryResponse.documents.front()["o"].Obj())) { - log() << "stopReplProducerOnDocument fail point is enabled."; - return Status(ErrorCodes::FailPointEnabled, - "stopReplProducerOnDocument fail point is enabled"); - } + { + Status status = Status::OK(); + stopReplProducerOnDocument.executeIf( + [&](auto&&) { + status = {ErrorCodes::FailPointEnabled, + "stopReplProducerOnDocument fail point is enabled."}; + log() << status.reason(); + }, + [&](const BSONObj& data) { + auto opCtx = cc().makeOperationContext(); + boost::intrusive_ptr<ExpressionContext> expCtx( + new ExpressionContext(opCtx.get(), nullptr)); + Matcher m(data["document"].Obj(), expCtx); + return !queryResponse.documents.empty() && + m.matches(queryResponse.documents.front()["o"].Obj()); + }); + if (!status.isOK()) + return status; } const auto& documents = queryResponse.documents; diff --git a/src/mongo/db/repl/oplog_fetcher.h b/src/mongo/db/repl/oplog_fetcher.h index f547304ac39..6bf7a0275eb 100644 --- a/src/mongo/db/repl/oplog_fetcher.h +++ b/src/mongo/db/repl/oplog_fetcher.h @@ -44,7 +44,7 @@ namespace mongo { namespace repl { -MONGO_FAIL_POINT_DECLARE(stopReplProducer); +extern FailPoint stopReplProducer; /** * The oplog fetcher, once started, reads operations from a remote oplog using a tailable cursor. diff --git a/src/mongo/db/repl/repl_set_commands.cpp b/src/mongo/db/repl/repl_set_commands.cpp index 096e74076a3..cadbce56e30 100644 --- a/src/mongo/db/repl/repl_set_commands.cpp +++ b/src/mongo/db/repl/repl_set_commands.cpp @@ -679,10 +679,8 @@ public: const string&, const BSONObj& cmdObj, BSONObjBuilder& result) { - MONGO_FAIL_POINT_BLOCK(rsDelayHeartbeatResponse, delay) { - const BSONObj& data = delay.getData(); - sleepsecs(data["delay"].numberInt()); - } + rsDelayHeartbeatResponse.execute( + [&](const BSONObj& data) { sleepsecs(data["delay"].numberInt()); }); LOG_FOR_HEARTBEATS(2) << "Received heartbeat request from " << cmdObj.getStringField("from") << ", " << cmdObj; diff --git a/src/mongo/db/repl/replication_coordinator_external_state_impl.cpp b/src/mongo/db/repl/replication_coordinator_external_state_impl.cpp index 596ece684c2..05806b9b9db 100644 --- a/src/mongo/db/repl/replication_coordinator_external_state_impl.cpp +++ b/src/mongo/db/repl/replication_coordinator_external_state_impl.cpp @@ -881,12 +881,12 @@ void ReplicationCoordinatorExternalStateImpl::notifyOplogMetadataWaiters( scheduleWork( _taskExecutor.get(), [committedOpTime, reaper](const executor::TaskExecutor::CallbackArgs& args) { - if (MONGO_FAIL_POINT(dropPendingCollectionReaperHang)) { + if (MONGO_unlikely(dropPendingCollectionReaperHang.shouldFail())) { log() << "fail point dropPendingCollectionReaperHang enabled. " "Blocking until fail point is disabled. " "committedOpTime: " << committedOpTime; - MONGO_FAIL_POINT_PAUSE_WHILE_SET(dropPendingCollectionReaperHang); + dropPendingCollectionReaperHang.pauseWhileSet(); } auto opCtx = cc().makeOperationContext(); reaper->dropCollectionsOlderThan(opCtx.get(), committedOpTime); diff --git a/src/mongo/db/repl/replication_coordinator_impl.cpp b/src/mongo/db/repl/replication_coordinator_impl.cpp index 5d34f1e691c..a98101fb33b 100644 --- a/src/mongo/db/repl/replication_coordinator_impl.cpp +++ b/src/mongo/db/repl/replication_coordinator_impl.cpp @@ -2030,11 +2030,12 @@ void ReplicationCoordinatorImpl::stepDown(OperationContext* opCtx, auto action = _updateMemberStateFromTopologyCoordinator(lk, opCtx); lk.unlock(); - if (MONGO_FAIL_POINT(stepdownHangBeforePerformingPostMemberStateUpdateActions)) { + if (MONGO_unlikely(stepdownHangBeforePerformingPostMemberStateUpdateActions.shouldFail())) { log() << "stepping down from primary - " "stepdownHangBeforePerformingPostMemberStateUpdateActions fail point enabled. " "Blocking until fail point is disabled."; - while (MONGO_FAIL_POINT(stepdownHangBeforePerformingPostMemberStateUpdateActions)) { + while (MONGO_unlikely( + stepdownHangBeforePerformingPostMemberStateUpdateActions.shouldFail())) { mongo::sleepsecs(1); { stdx::lock_guard<stdx::mutex> lock(_mutex); @@ -3567,11 +3568,10 @@ boost::optional<OpTimeAndWallTime> ReplicationCoordinatorImpl::_chooseStableOpTi maximumStableOpTime.opTime.getTimestamp()); } - MONGO_FAIL_POINT_BLOCK(holdStableTimestampAtSpecificTimestamp, data) { - const BSONObj& dataObj = data.getData(); + holdStableTimestampAtSpecificTimestamp.execute([&](const BSONObj& dataObj) { const auto holdStableTimestamp = dataObj["timestamp"].timestamp(); maximumStableTimestamp = std::min(maximumStableTimestamp, holdStableTimestamp); - } + }); maximumStableOpTime = {OpTime(maximumStableTimestamp, maximumStableOpTime.opTime.getTerm()), maximumStableOpTime.wallTime}; @@ -3693,7 +3693,7 @@ void ReplicationCoordinatorImpl::_setStableTimestampForStorage(WithLock lk) { } // Set the stable timestamp regardless of whether the majority commit point moved // forward. - if (!MONGO_FAIL_POINT(disableSnapshotting)) { + if (!MONGO_unlikely(disableSnapshotting.shouldFail())) { _storage->setStableTimestamp(getServiceContext(), stableOpTime->opTime.getTimestamp()); } @@ -3980,7 +3980,7 @@ bool ReplicationCoordinatorImpl::_updateCommittedSnapshot( _currentCommittedSnapshot->opTime.getTimestamp()); invariant(newCommittedSnapshot.opTime >= _currentCommittedSnapshot->opTime); } - if (MONGO_FAIL_POINT(disableSnapshotting)) + if (MONGO_unlikely(disableSnapshotting.shouldFail())) return false; _currentCommittedSnapshot = newCommittedSnapshot; _currentCommittedSnapshotCond.notify_all(); diff --git a/src/mongo/db/repl/replication_coordinator_impl_elect_v1.cpp b/src/mongo/db/repl/replication_coordinator_impl_elect_v1.cpp index b1464c0de9b..f49ecec21be 100644 --- a/src/mongo/db/repl/replication_coordinator_impl_elect_v1.cpp +++ b/src/mongo/db/repl/replication_coordinator_impl_elect_v1.cpp @@ -355,13 +355,13 @@ void ReplicationCoordinatorImpl::_onVoteRequestComplete( _voteRequester.reset(); auto electionFinishedEvent = _electionFinishedEvent; - MONGO_FAIL_POINT_BLOCK(electionHangsBeforeUpdateMemberState, customWait) { - auto waitForMillis = Milliseconds(customWait.getData()["waitForMillis"].numberInt()); + electionHangsBeforeUpdateMemberState.execute([&](const BSONObj& customWait) { + auto waitForMillis = Milliseconds(customWait["waitForMillis"].numberInt()); log() << "election succeeded - electionHangsBeforeUpdateMemberState fail point " "enabled, sleeping " << waitForMillis; sleepFor(waitForMillis); - } + }); _postWonElectionUpdateMemberState(lk); _replExecutor->signalEvent(electionFinishedEvent); diff --git a/src/mongo/db/repl/replication_coordinator_impl_elect_v1_test.cpp b/src/mongo/db/repl/replication_coordinator_impl_elect_v1_test.cpp index 4c7afc1130f..c51f4ace311 100644 --- a/src/mongo/db/repl/replication_coordinator_impl_elect_v1_test.cpp +++ b/src/mongo/db/repl/replication_coordinator_impl_elect_v1_test.cpp @@ -478,9 +478,7 @@ TEST_F(ReplCoordTest, NodeWillNotStandForElectionDuringHeartbeatReconfig) { replCoordSetMyLastAppliedOpTime(OpTime(Timestamp(100, 1), 0), Date_t() + Seconds(100)); replCoordSetMyLastDurableOpTime(OpTime(Timestamp(100, 1), 0), Date_t() + Seconds(100)); - getGlobalFailPointRegistry() - ->getFailPoint("blockHeartbeatReconfigFinish") - ->setMode(FailPoint::alwaysOn); + globalFailPointRegistry().find("blockHeartbeatReconfigFinish")->setMode(FailPoint::alwaysOn); // hb reconfig NetworkInterfaceMock* net = getNet(); @@ -573,9 +571,7 @@ TEST_F(ReplCoordTest, NodeWillNotStandForElectionDuringHeartbeatReconfig) { ASSERT_EQUALS(1, countLogLinesContaining("Not standing for election; processing " "a configuration change")); - getGlobalFailPointRegistry() - ->getFailPoint("blockHeartbeatReconfigFinish") - ->setMode(FailPoint::off); + globalFailPointRegistry().find("blockHeartbeatReconfigFinish")->setMode(FailPoint::off); } TEST_F(ReplCoordTest, ElectionFailsWhenInsufficientVotesAreReceivedDuringRequestVotes) { diff --git a/src/mongo/db/repl/replication_coordinator_impl_heartbeat.cpp b/src/mongo/db/repl/replication_coordinator_impl_heartbeat.cpp index 8900e2a0471..6097df4f6e0 100644 --- a/src/mongo/db/repl/replication_coordinator_impl_heartbeat.cpp +++ b/src/mongo/db/repl/replication_coordinator_impl_heartbeat.cpp @@ -370,7 +370,7 @@ void ReplicationCoordinatorImpl::_stepDownFinish( return; } - if (MONGO_FAIL_POINT(blockHeartbeatStepdown)) { + if (MONGO_unlikely(blockHeartbeatStepdown.shouldFail())) { // This log output is used in js tests so please leave it. log() << "stepDown - blockHeartbeatStepdown fail point enabled. " "Blocking until fail point is disabled."; @@ -380,7 +380,7 @@ void ReplicationCoordinatorImpl::_stepDownFinish( return _inShutdown; }; - while (MONGO_FAIL_POINT(blockHeartbeatStepdown) && !inShutdown()) { + while (MONGO_unlikely(blockHeartbeatStepdown.shouldFail()) && !inShutdown()) { mongo::sleepsecs(1); } } @@ -577,7 +577,7 @@ void ReplicationCoordinatorImpl::_heartbeatReconfigFinish( return; } - if (MONGO_FAIL_POINT(blockHeartbeatReconfigFinish)) { + if (MONGO_unlikely(blockHeartbeatReconfigFinish.shouldFail())) { LOG_FOR_HEARTBEATS(0) << "blockHeartbeatReconfigFinish fail point enabled. Rescheduling " "_heartbeatReconfigFinish until fail point is disabled."; _replExecutor diff --git a/src/mongo/db/repl/replication_coordinator_impl_reconfig_test.cpp b/src/mongo/db/repl/replication_coordinator_impl_reconfig_test.cpp index 738ff86ef87..e90146ecf34 100644 --- a/src/mongo/db/repl/replication_coordinator_impl_reconfig_test.cpp +++ b/src/mongo/db/repl/replication_coordinator_impl_reconfig_test.cpp @@ -469,9 +469,7 @@ TEST_F( simulateSuccessfulV1Election(); ASSERT_TRUE(getReplCoord()->getMemberState().primary()); - getGlobalFailPointRegistry() - ->getFailPoint("blockHeartbeatReconfigFinish") - ->setMode(FailPoint::alwaysOn); + globalFailPointRegistry().find("blockHeartbeatReconfigFinish")->setMode(FailPoint::alwaysOn); // hb reconfig NetworkInterfaceMock* net = getNet(); @@ -511,9 +509,7 @@ TEST_F( ASSERT_EQUALS(ErrorCodes::ConfigurationInProgress, getReplCoord()->processReplSetReconfig(opCtx.get(), args, &result)); - getGlobalFailPointRegistry() - ->getFailPoint("blockHeartbeatReconfigFinish") - ->setMode(FailPoint::off); + globalFailPointRegistry().find("blockHeartbeatReconfigFinish")->setMode(FailPoint::off); } TEST_F(ReplCoordTest, NodeDoesNotAcceptHeartbeatReconfigWhileInTheMidstOfReconfig) { diff --git a/src/mongo/db/repl/replication_coordinator_impl_test.cpp b/src/mongo/db/repl/replication_coordinator_impl_test.cpp index 631a47c6395..117f76eed7d 100644 --- a/src/mongo/db/repl/replication_coordinator_impl_test.cpp +++ b/src/mongo/db/repl/replication_coordinator_impl_test.cpp @@ -137,9 +137,7 @@ TEST_F(ReplCoordTest, IsMasterIsFalseDuringStepdown) { ASSERT(replCoord->getMemberState().primary()); // Primary begins stepping down due to new term, but cannot finish. - getGlobalFailPointRegistry() - ->getFailPoint("blockHeartbeatStepdown") - ->setMode(FailPoint::alwaysOn); + globalFailPointRegistry().find("blockHeartbeatStepdown")->setMode(FailPoint::alwaysOn); TopologyCoordinator::UpdateTermResult updateTermResult; replCoord->updateTerm_forTest(replCoord->getTerm() + 1, &updateTermResult); @@ -154,7 +152,7 @@ TEST_F(ReplCoordTest, IsMasterIsFalseDuringStepdown) { ASSERT_FALSE(responseObj["secondary"].Bool()); ASSERT_FALSE(responseObj.hasField("isreplicaset")); - getGlobalFailPointRegistry()->getFailPoint("blockHeartbeatStepdown")->setMode(FailPoint::off); + globalFailPointRegistry().find("blockHeartbeatStepdown")->setMode(FailPoint::off); } TEST_F(ReplCoordTest, NodeEntersStartup2StateWhenStartingUpWithValidLocalConfig) { diff --git a/src/mongo/db/repl/replication_coordinator_test_fixture.cpp b/src/mongo/db/repl/replication_coordinator_test_fixture.cpp index a2e5e2d9e8a..27fac4a3715 100644 --- a/src/mongo/db/repl/replication_coordinator_test_fixture.cpp +++ b/src/mongo/db/repl/replication_coordinator_test_fixture.cpp @@ -86,9 +86,7 @@ ReplCoordTest::ReplCoordTest() { } ReplCoordTest::~ReplCoordTest() { - getGlobalFailPointRegistry() - ->getFailPoint("blockHeartbeatReconfigFinish") - ->setMode(FailPoint::off); + globalFailPointRegistry().find("blockHeartbeatReconfigFinish")->setMode(FailPoint::off); if (_callShutdown) { auto opCtx = makeOperationContext(); diff --git a/src/mongo/db/repl/replication_info.cpp b/src/mongo/db/repl/replication_info.cpp index 6e93691e17b..9b854735676 100644 --- a/src/mongo/db/repl/replication_info.cpp +++ b/src/mongo/db/repl/replication_info.cpp @@ -244,7 +244,7 @@ public: CommandHelpers::handleMarkKillOnClientDisconnect(opCtx); // TODO Unwind after SERVER-41070 - MONGO_FAIL_POINT_PAUSE_WHILE_SET_OR_INTERRUPTED(opCtx, waitInIsMaster); + waitInIsMaster.pauseWhileSet(opCtx); /* currently request to arbiter is (somewhat arbitrarily) an ismaster request that is not authenticated. diff --git a/src/mongo/db/repl/roll_back_local_operations.h b/src/mongo/db/repl/roll_back_local_operations.h index 6c6e6c148ab..b6920cc10c4 100644 --- a/src/mongo/db/repl/roll_back_local_operations.h +++ b/src/mongo/db/repl/roll_back_local_operations.h @@ -48,11 +48,11 @@ namespace repl { // two separate files, rs_rollback and rs_rollback_no_uuid. However, after // MongoDB 3.8 is released, we no longer need to maintain rs_rollback_no_uuid // code and these forward declares can be removed. See SERVER-29766. -MONGO_FAIL_POINT_DECLARE(rollbackHangBeforeFinish); -MONGO_FAIL_POINT_DECLARE(rollbackHangThenFailAfterWritingMinValid); +extern FailPoint rollbackHangBeforeFinish; +extern FailPoint rollbackHangThenFailAfterWritingMinValid; // This is needed by rs_rollback and rollback_impl. -MONGO_FAIL_POINT_DECLARE(rollbackHangAfterTransitionToRollback); +extern FailPoint rollbackHangAfterTransitionToRollback; class RollBackLocalOperations { RollBackLocalOperations(const RollBackLocalOperations&) = delete; diff --git a/src/mongo/db/repl/rollback_impl.cpp b/src/mongo/db/repl/rollback_impl.cpp index 1dc03162c29..ecb73b66573 100644 --- a/src/mongo/db/repl/rollback_impl.cpp +++ b/src/mongo/db/repl/rollback_impl.cpp @@ -182,11 +182,10 @@ Status RollbackImpl::runRollback(OperationContext* opCtx) { } _listener->onTransitionToRollback(); - if (MONGO_FAIL_POINT(rollbackHangAfterTransitionToRollback)) { + if (MONGO_unlikely(rollbackHangAfterTransitionToRollback.shouldFail())) { log() << "rollbackHangAfterTransitionToRollback fail point enabled. Blocking until fail " "point is disabled (rollback_impl)."; - MONGO_FAIL_POINT_PAUSE_WHILE_SET_OR_INTERRUPTED(opCtx, - rollbackHangAfterTransitionToRollback); + rollbackHangAfterTransitionToRollback.pauseWhileSet(opCtx); } // We clear the SizeRecoveryState before we recover to a stable timestamp. This ensures that we diff --git a/src/mongo/db/repl/rs_rollback.cpp b/src/mongo/db/repl/rs_rollback.cpp index 04ee064da24..3e50d1ede2d 100644 --- a/src/mongo/db/repl/rs_rollback.cpp +++ b/src/mongo/db/repl/rs_rollback.cpp @@ -670,12 +670,12 @@ void checkRbidAndUpdateMinValid(OperationContext* opCtx, replicationProcess->getConsistencyMarkers()->clearAppliedThrough(opCtx, {}); replicationProcess->getConsistencyMarkers()->setMinValid(opCtx, minValid); - if (MONGO_FAIL_POINT(rollbackHangThenFailAfterWritingMinValid)) { + if (MONGO_unlikely(rollbackHangThenFailAfterWritingMinValid.shouldFail())) { // This log output is used in jstests so please leave it. log() << "rollback - rollbackHangThenFailAfterWritingMinValid fail point " "enabled. Blocking until fail point is disabled."; - while (MONGO_FAIL_POINT(rollbackHangThenFailAfterWritingMinValid)) { + while (MONGO_unlikely(rollbackHangThenFailAfterWritingMinValid.shouldFail())) { invariant(!globalInShutdownDeprecated()); // It is an error to shutdown while enabled. mongo::sleepsecs(1); } @@ -1001,7 +1001,7 @@ Status _syncRollback(OperationContext* opCtx, }); syncFixUp(opCtx, how, rollbackSource, replCoord, replicationProcess); - if (MONGO_FAIL_POINT(rollbackExitEarlyAfterCollectionDrop)) { + if (MONGO_unlikely(rollbackExitEarlyAfterCollectionDrop.shouldFail())) { log() << "rollbackExitEarlyAfterCollectionDrop fail point enabled. Returning early " "until fail point is disabled."; return Status(ErrorCodes::NamespaceNotFound, @@ -1013,11 +1013,11 @@ Status _syncRollback(OperationContext* opCtx, return Status(ErrorCodes::UnrecoverableRollbackError, e.what()); } - if (MONGO_FAIL_POINT(rollbackHangBeforeFinish)) { + if (MONGO_unlikely(rollbackHangBeforeFinish.shouldFail())) { // This log output is used in js tests so please leave it. log() << "rollback - rollbackHangBeforeFinish fail point " "enabled. Blocking until fail point is disabled."; - while (MONGO_FAIL_POINT(rollbackHangBeforeFinish)) { + while (MONGO_unlikely(rollbackHangBeforeFinish.shouldFail())) { invariant(!globalInShutdownDeprecated()); // It is an error to shutdown while enabled. mongo::sleepsecs(1); } @@ -1181,7 +1181,7 @@ void rollback_internal::syncFixUp(OperationContext* opCtx, } } - if (MONGO_FAIL_POINT(rollbackExitEarlyAfterCollectionDrop)) { + if (MONGO_unlikely(rollbackExitEarlyAfterCollectionDrop.shouldFail())) { return; } @@ -1643,11 +1643,10 @@ void rollback(OperationContext* opCtx, } } - if (MONGO_FAIL_POINT(rollbackHangAfterTransitionToRollback)) { + if (MONGO_unlikely(rollbackHangAfterTransitionToRollback.shouldFail())) { log() << "rollbackHangAfterTransitionToRollback fail point enabled. Blocking until fail " "point is disabled (rs_rollback)."; - MONGO_FAIL_POINT_PAUSE_WHILE_SET_OR_INTERRUPTED(opCtx, - rollbackHangAfterTransitionToRollback); + rollbackHangAfterTransitionToRollback.pauseWhileSet(opCtx); } try { diff --git a/src/mongo/db/repl/sync_tail.cpp b/src/mongo/db/repl/sync_tail.cpp index ec30792335f..165ab31c7a3 100644 --- a/src/mongo/db/repl/sync_tail.cpp +++ b/src/mongo/db/repl/sync_tail.cpp @@ -530,7 +530,7 @@ private: BatchLimits batchLimits; while (true) { - MONGO_FAIL_POINT_PAUSE_WHILE_SET(rsSyncApplyStop); + rsSyncApplyStop.pauseWhileSet(); batchLimits.slaveDelayLatestTimestamp = _calculateSlaveDelayLatestTimestamp(); @@ -661,10 +661,10 @@ void SyncTail::_oplogApplication(ReplicationCoordinator* replCoord, opCtx.setShouldParticipateInFlowControl(false); // For pausing replication in tests. - if (MONGO_FAIL_POINT(rsSyncApplyStop)) { + if (MONGO_unlikely(rsSyncApplyStop.shouldFail())) { log() << "sync tail - rsSyncApplyStop fail point enabled. Blocking until fail point is " "disabled."; - while (MONGO_FAIL_POINT(rsSyncApplyStop)) { + while (MONGO_unlikely(rsSyncApplyStop.shouldFail())) { // Tests should not trigger clean shutdown while that failpoint is active. If we // think we need this, we need to think hard about what the behavior should be. if (inShutdown()) { @@ -689,7 +689,7 @@ void SyncTail::_oplogApplication(ReplicationCoordinator* replCoord, // Shut down and exit oplog application loop. return; } - if (MONGO_FAIL_POINT(rsSyncApplyStop)) { + if (MONGO_unlikely(rsSyncApplyStop.shouldFail())) { continue; } if (ops.termWhenExhausted()) { @@ -814,10 +814,10 @@ Status syncApply(OperationContext* opCtx, auto clockSource = opCtx->getServiceContext()->getFastClockSource(); auto applyStartTime = clockSource->now(); - if (MONGO_FAIL_POINT(hangAfterRecordingOpApplicationStartTime)) { + if (MONGO_unlikely(hangAfterRecordingOpApplicationStartTime.shouldFail())) { log() << "syncApply - fail point hangAfterRecordingOpApplicationStartTime enabled. " << "Blocking until fail point is disabled. "; - MONGO_FAIL_POINT_PAUSE_WHILE_SET(hangAfterRecordingOpApplicationStartTime); + hangAfterRecordingOpApplicationStartTime.pauseWhileSet(); } auto opType = op.getOpType(); @@ -1198,11 +1198,10 @@ StatusWith<OpTime> SyncTail::multiApply(OperationContext* opCtx, MultiApplier::O // Use this fail point to hold the PBWM lock after we have written the oplog entries but // before we have applied them. - if (MONGO_FAIL_POINT(pauseBatchApplicationAfterWritingOplogEntries)) { + if (MONGO_unlikely(pauseBatchApplicationAfterWritingOplogEntries.shouldFail())) { log() << "pauseBatchApplicationAfterWritingOplogEntries fail point enabled. Blocking " "until fail point is disabled."; - MONGO_FAIL_POINT_PAUSE_WHILE_SET_OR_INTERRUPTED( - opCtx, pauseBatchApplicationAfterWritingOplogEntries); + pauseBatchApplicationAfterWritingOplogEntries.pauseWhileSet(opCtx); } // Reset consistency markers in case the node fails while applying ops. @@ -1239,10 +1238,10 @@ StatusWith<OpTime> SyncTail::multiApply(OperationContext* opCtx, MultiApplier::O storageEngine->replicationBatchIsComplete(); // Use this fail point to hold the PBWM lock and prevent the batch from completing. - if (MONGO_FAIL_POINT(pauseBatchApplicationBeforeCompletion)) { + if (MONGO_unlikely(pauseBatchApplicationBeforeCompletion.shouldFail())) { log() << "pauseBatchApplicationBeforeCompletion fail point enabled. Blocking until fail " "point is disabled."; - while (MONGO_FAIL_POINT(pauseBatchApplicationBeforeCompletion)) { + while (MONGO_unlikely(pauseBatchApplicationBeforeCompletion.shouldFail())) { if (inShutdown()) { severe() << "Turn off pauseBatchApplicationBeforeCompletion before attempting " "clean shutdown"; diff --git a/src/mongo/db/repl/topology_coordinator.cpp b/src/mongo/db/repl/topology_coordinator.cpp index c79248d7af6..1441a4cdeb5 100644 --- a/src/mongo/db/repl/topology_coordinator.cpp +++ b/src/mongo/db/repl/topology_coordinator.cpp @@ -208,8 +208,8 @@ HostAndPort TopologyCoordinator::chooseNewSyncSource(Date_t now, return HostAndPort(); } - MONGO_FAIL_POINT_BLOCK(forceSyncSourceCandidate, customArgs) { - const auto& data = customArgs.getData(); + if (auto sfp = forceSyncSourceCandidate.scoped(); MONGO_unlikely(sfp.isActive())) { + const auto& data = sfp.getData(); const auto hostAndPortElem = data["hostAndPort"]; if (!hostAndPortElem) { severe() << "'forceSyncSoureCandidate' parameter set with invalid host and port: " @@ -2659,7 +2659,7 @@ bool TopologyCoordinator::shouldChangeSyncSource( return true; } - if (MONGO_FAIL_POINT(disableMaxSyncSourceLagSecs)) { + if (MONGO_unlikely(disableMaxSyncSourceLagSecs.shouldFail())) { log() << "disableMaxSyncSourceLagSecs fail point enabled - not checking the most recent " "OpTime, " << currentSourceOpTime.toString() << ", of our current sync source, " << currentSource diff --git a/src/mongo/db/repl/transaction_oplog_application.cpp b/src/mongo/db/repl/transaction_oplog_application.cpp index 92cbee05379..2d65483c42c 100644 --- a/src/mongo/db/repl/transaction_oplog_application.cpp +++ b/src/mongo/db/repl/transaction_oplog_application.cpp @@ -355,10 +355,9 @@ Status _applyPrepareTransaction(OperationContext* opCtx, auto status = _applyOperationsForTransaction(opCtx, ops, mode); fassert(31137, status); - if (MONGO_FAIL_POINT(applyOpsHangBeforePreparingTransaction)) { + if (MONGO_unlikely(applyOpsHangBeforePreparingTransaction.shouldFail())) { LOG(0) << "Hit applyOpsHangBeforePreparingTransaction failpoint"; - MONGO_FAIL_POINT_PAUSE_WHILE_SET_OR_INTERRUPTED(opCtx, - applyOpsHangBeforePreparingTransaction); + applyOpsHangBeforePreparingTransaction.pauseWhileSet(opCtx); } transaction.prepareTransaction(opCtx, entry.getOpTime()); @@ -435,7 +434,7 @@ Status applyPrepareTransaction(OperationContext* opCtx, } void reconstructPreparedTransactions(OperationContext* opCtx, repl::OplogApplication::Mode mode) { - if (MONGO_FAIL_POINT(skipReconstructPreparedTransactions)) { + if (MONGO_unlikely(skipReconstructPreparedTransactions.shouldFail())) { log() << "Hit skipReconstructPreparedTransactions failpoint"; return; } diff --git a/src/mongo/db/s/balancer/balancer.cpp b/src/mongo/db/s/balancer/balancer.cpp index dd1a9011612..a920dd64ba9 100644 --- a/src/mongo/db/s/balancer/balancer.cpp +++ b/src/mongo/db/s/balancer/balancer.cpp @@ -394,18 +394,14 @@ void Balancer::_mainThread() { LOG(1) << "*** End of balancing round"; } - auto balancerInterval = [&]() -> Milliseconds { - MONGO_FAIL_POINT_BLOCK(overrideBalanceRoundInterval, data) { - int interval = data.getData()["intervalMs"].numberInt(); - log() << "overrideBalanceRoundInterval: using shorter balancing interval: " - << interval << "ms"; - - return Milliseconds(interval); - } - - return _balancedLastTime ? kShortBalanceRoundInterval - : kBalanceRoundDefaultInterval; - }(); + Milliseconds balancerInterval = + _balancedLastTime ? kShortBalanceRoundInterval : kBalanceRoundDefaultInterval; + + overrideBalanceRoundInterval.execute([&](const BSONObj& data) { + balancerInterval = Milliseconds(data["intervalMs"].numberInt()); + log() << "overrideBalanceRoundInterval: using shorter balancing interval: " + << balancerInterval; + }); _endRound(opCtx.get(), balancerInterval); } catch (const DBException& e) { diff --git a/src/mongo/db/s/balancer/balancer_policy.cpp b/src/mongo/db/s/balancer/balancer_policy.cpp index efd71ca26ed..f512f77ca6a 100644 --- a/src/mongo/db/s/balancer/balancer_policy.cpp +++ b/src/mongo/db/s/balancer/balancer_policy.cpp @@ -355,7 +355,7 @@ vector<MigrateInfo> BalancerPolicy::balance(const ShardStatisticsVector& shardSt std::set<ShardId>* usedShards) { vector<MigrateInfo> migrations; - if (MONGO_FAIL_POINT(balancerShouldReturnRandomMigrations) && + if (MONGO_unlikely(balancerShouldReturnRandomMigrations.shouldFail()) && !distribution.nss().isConfigDB()) { LOG(1) << "balancerShouldReturnRandomMigrations failpoint is set"; diff --git a/src/mongo/db/s/collection_range_deleter.cpp b/src/mongo/db/s/collection_range_deleter.cpp index d5affc26cc0..27b0a47a7ef 100644 --- a/src/mongo/db/s/collection_range_deleter.cpp +++ b/src/mongo/db/s/collection_range_deleter.cpp @@ -395,9 +395,9 @@ StatusWith<int> CollectionRangeDeleter::_doDeletion(OperationContext* opCtx, PlanExecutor::YIELD_MANUAL, InternalPlanner::FORWARD); - if (MONGO_FAIL_POINT(hangBeforeDoingDeletion)) { + if (MONGO_unlikely(hangBeforeDoingDeletion.shouldFail())) { LOG(0) << "Hit hangBeforeDoingDeletion failpoint"; - MONGO_FAIL_POINT_PAUSE_WHILE_SET_OR_INTERRUPTED(opCtx, hangBeforeDoingDeletion); + hangBeforeDoingDeletion.pauseWhileSet(opCtx); } PlanYieldPolicy planYieldPolicy(exec.get(), PlanExecutor::YIELD_MANUAL); diff --git a/src/mongo/db/s/collection_sharding_state.cpp b/src/mongo/db/s/collection_sharding_state.cpp index dbe03cadd4d..7e38463b55e 100644 --- a/src/mongo/db/s/collection_sharding_state.cpp +++ b/src/mongo/db/s/collection_sharding_state.cpp @@ -226,7 +226,7 @@ boost::optional<ScopedCollectionMetadata> CollectionShardingState::_getMetadataW auto metadata = _getMetadata(atClusterTime); auto wantedShardVersion = ChunkVersion::UNSHARDED(); - if (MONGO_FAIL_POINT(useFCV44CheckShardVersionProtocol)) { + if (MONGO_unlikely(useFCV44CheckShardVersionProtocol.shouldFail())) { LOG(0) << "Received shardVersion: " << receivedShardVersion << " for " << _nss.ns(); if (isCollection) { LOG(0) << "Namespace " << _nss.ns() << " is collection, " diff --git a/src/mongo/db/s/config/configsvr_drop_collection_command.cpp b/src/mongo/db/s/config/configsvr_drop_collection_command.cpp index 66c2cb94653..06970e7ecf3 100644 --- a/src/mongo/db/s/config/configsvr_drop_collection_command.cpp +++ b/src/mongo/db/s/config/configsvr_drop_collection_command.cpp @@ -111,10 +111,8 @@ public: opCtx->getWriteConcern().wMode == WriteConcernOptions::kMajority); Seconds waitFor(DistLockManager::kDefaultLockTimeout); - MONGO_FAIL_POINT_BLOCK(setDropCollDistLockWait, customWait) { - const BSONObj& data = customWait.getData(); - waitFor = Seconds(data["waitForSecs"].numberInt()); - } + setDropCollDistLockWait.execute( + [&](const BSONObj& data) { waitFor = Seconds(data["waitForSecs"].numberInt()); }); auto const catalogClient = Grid::get(opCtx)->catalogClient(); diff --git a/src/mongo/db/s/config/sharding_catalog_manager_add_shard_test.cpp b/src/mongo/db/s/config/sharding_catalog_manager_add_shard_test.cpp index 337f854354e..00287e86410 100644 --- a/src/mongo/db/s/config/sharding_catalog_manager_add_shard_test.cpp +++ b/src/mongo/db/s/config/sharding_catalog_manager_add_shard_test.cpp @@ -997,7 +997,7 @@ TEST_F(AddShardTest, AddShardSucceedsEvenIfAddingDBsFromNewShardFails) { // Enable fail point to cause all updates to fail. Since we add the databases detected from // the shard being added with upserts, but we add the shard document itself via insert, this // will allow the shard to be added but prevent the databases from brought into the cluster. - auto failPoint = getGlobalFailPointRegistry()->getFailPoint("failAllUpdates"); + auto failPoint = globalFailPointRegistry().find("failAllUpdates"); ASSERT(failPoint); failPoint->setMode(FailPoint::alwaysOn); ON_BLOCK_EXIT([&] { failPoint->setMode(FailPoint::off); }); diff --git a/src/mongo/db/s/config/sharding_catalog_manager_chunk_operations.cpp b/src/mongo/db/s/config/sharding_catalog_manager_chunk_operations.cpp index e3683f37787..76df6923276 100644 --- a/src/mongo/db/s/config/sharding_catalog_manager_chunk_operations.cpp +++ b/src/mongo/db/s/config/sharding_catalog_manager_chunk_operations.cpp @@ -663,7 +663,7 @@ StatusWith<BSONObj> ShardingCatalogManager::commitChunkMigration( return findResponse.getStatus(); } - if (MONGO_FAIL_POINT(migrationCommitVersionError)) { + if (MONGO_unlikely(migrationCommitVersionError.shouldFail())) { uassert(ErrorCodes::StaleEpoch, "failpoint 'migrationCommitVersionError' generated error", false); @@ -730,7 +730,7 @@ StatusWith<BSONObj> ShardingCatalogManager::commitChunkMigration( // Update the history of the migrated chunk. // Drop the history that is too old (10 seconds of history for now). // TODO SERVER-33831 to update the old history removal policy. - if (!MONGO_FAIL_POINT(skipExpiringOldChunkHistory)) { + if (!MONGO_unlikely(skipExpiringOldChunkHistory.shouldFail())) { while (!newHistory.empty() && newHistory.back().getValidAfter().getSecs() + kHistorySecs < validAfter.get().getSecs()) { diff --git a/src/mongo/db/s/config/sharding_catalog_manager_collection_operations.cpp b/src/mongo/db/s/config/sharding_catalog_manager_collection_operations.cpp index cba7c79772e..28122a3a3da 100644 --- a/src/mongo/db/s/config/sharding_catalog_manager_collection_operations.cpp +++ b/src/mongo/db/s/config/sharding_catalog_manager_collection_operations.cpp @@ -527,10 +527,9 @@ void ShardingCatalogManager::renameCollection(OperationContext* opCtx, CommandHelpers::filterCommandRequestForPassthrough(passthroughFields)), Shard::RetryPolicy::kIdempotent)); - if (MONGO_FAIL_POINT(hangRenameCollectionAfterSendingRenameToPrimaryShard)) { + if (MONGO_unlikely(hangRenameCollectionAfterSendingRenameToPrimaryShard.shouldFail())) { log() << "Hit hangRenameCollectionAfterSendingRenameToPrimaryShard"; - MONGO_FAIL_POINT_PAUSE_WHILE_SET_OR_INTERRUPTED( - opCtx, hangRenameCollectionAfterSendingRenameToPrimaryShard); + hangRenameCollectionAfterSendingRenameToPrimaryShard.pauseWhileSet(opCtx); } uassertStatusOK(cmdResponse.commandStatus); @@ -638,10 +637,9 @@ void ShardingCatalogManager::generateUUIDsForExistingShardedCollections(Operatio void ShardingCatalogManager::createCollection(OperationContext* opCtx, const NamespaceString& ns, const CollectionOptions& collOptions) { - if (MONGO_FAIL_POINT(hangCreateCollectionAfterAcquiringDistlocks)) { + if (MONGO_unlikely(hangCreateCollectionAfterAcquiringDistlocks.shouldFail())) { log() << "Hit hangCreateCollectionAfterAcquiringDistlocks"; - MONGO_FAIL_POINT_PAUSE_WHILE_SET_OR_INTERRUPTED( - opCtx, hangCreateCollectionAfterAcquiringDistlocks); + hangCreateCollectionAfterAcquiringDistlocks.pauseWhileSet(opCtx); } const auto catalogClient = Grid::get(opCtx)->catalogClient(); @@ -668,10 +666,9 @@ void ShardingCatalogManager::createCollection(OperationContext* opCtx, createCmdBuilder.obj(), Shard::RetryPolicy::kIdempotent); - if (MONGO_FAIL_POINT(hangCreateCollectionAfterSendingCreateToPrimaryShard)) { + if (MONGO_unlikely(hangCreateCollectionAfterSendingCreateToPrimaryShard.shouldFail())) { log() << "Hit hangCreateCollectionAfterSendingCreateToPrimaryShard"; - MONGO_FAIL_POINT_PAUSE_WHILE_SET_OR_INTERRUPTED( - opCtx, hangCreateCollectionAfterSendingCreateToPrimaryShard); + hangCreateCollectionAfterSendingCreateToPrimaryShard.pauseWhileSet(opCtx); } auto createStatus = Shard::CommandResponse::getEffectiveStatus(swResponse); @@ -681,10 +678,9 @@ void ShardingCatalogManager::createCollection(OperationContext* opCtx, const auto uuid = checkCollectionOptions(opCtx, primaryShard.get(), ns, collOptions); - if (MONGO_FAIL_POINT(hangCreateCollectionAfterGettingUUIDFromPrimaryShard)) { + if (MONGO_unlikely(hangCreateCollectionAfterGettingUUIDFromPrimaryShard.shouldFail())) { log() << "Hit hangCreateCollectionAfterGettingUUIDFromPrimaryShard"; - MONGO_FAIL_POINT_PAUSE_WHILE_SET_OR_INTERRUPTED( - opCtx, hangCreateCollectionAfterGettingUUIDFromPrimaryShard); + hangCreateCollectionAfterGettingUUIDFromPrimaryShard.pauseWhileSet(opCtx); } if (collOptions.isView()) { @@ -751,15 +747,14 @@ void ShardingCatalogManager::createCollection(OperationContext* opCtx, ShardingCatalogClient::kLocalWriteConcern)); } - if (MONGO_FAIL_POINT(writeUnshardedCollectionsToShardingCatalog)) { + if (MONGO_unlikely(writeUnshardedCollectionsToShardingCatalog.shouldFail())) { LOG(0) << "Going to write initial chunk for new unsharded collection " << ns.ns() << ": " << chunk.toString(); writeFirstChunksForCollection(opCtx, initialChunks); - if (MONGO_FAIL_POINT(hangCreateCollectionAfterWritingEntryToConfigChunks)) { + if (MONGO_unlikely(hangCreateCollectionAfterWritingEntryToConfigChunks.shouldFail())) { log() << "Hit hangCreateCollectionAfterWritingEntryToConfigChunks"; - MONGO_FAIL_POINT_PAUSE_WHILE_SET_OR_INTERRUPTED( - opCtx, hangCreateCollectionAfterWritingEntryToConfigChunks); + hangCreateCollectionAfterWritingEntryToConfigChunks.pauseWhileSet(opCtx); } LOG(0) << "Going to write collection entry for new unsharded collection " << ns.ns() << ": " @@ -767,10 +762,9 @@ void ShardingCatalogManager::createCollection(OperationContext* opCtx, uassertStatusOK(ShardingCatalogClientImpl::updateShardingCatalogEntryForCollection( opCtx, ns, targetCollType, true /*upsert*/)); - if (MONGO_FAIL_POINT(hangCreateCollectionAfterWritingEntryToConfigCollections)) { + if (MONGO_unlikely(hangCreateCollectionAfterWritingEntryToConfigCollections.shouldFail())) { log() << "Hit hangCreateCollectionAfterWritingEntryToConfigCollections"; - MONGO_FAIL_POINT_PAUSE_WHILE_SET_OR_INTERRUPTED( - opCtx, hangCreateCollectionAfterWritingEntryToConfigCollections); + hangCreateCollectionAfterWritingEntryToConfigCollections.pauseWhileSet(opCtx); } } } @@ -859,10 +853,9 @@ void ShardingCatalogManager::refineCollectionShardKey(OperationContext* opCtx, // the newly-generated objectid, (ii) their bounds for each new field in the refined key to // MinKey (except for the global max chunk where the max bounds are set to MaxKey), and // unsetting (iii) their jumbo field. - if (MONGO_FAIL_POINT(hangRefineCollectionShardKeyBeforeUpdatingChunks)) { + if (MONGO_unlikely(hangRefineCollectionShardKeyBeforeUpdatingChunks.shouldFail())) { log() << "Hit hangRefineCollectionShardKeyBeforeUpdatingChunks failpoint"; - MONGO_FAIL_POINT_PAUSE_WHILE_SET_OR_INTERRUPTED( - opCtx, hangRefineCollectionShardKeyBeforeUpdatingChunks); + hangRefineCollectionShardKeyBeforeUpdatingChunks.pauseWhileSet(opCtx); } uassertStatusOK(updateConfigDocumentInTxn(asr.opCtx(), @@ -918,10 +911,9 @@ void ShardingCatalogManager::refineCollectionShardKey(OperationContext* opCtx, << executionTimer.millis() << " ms. Total time taken: " << totalTimer.millis() << " ms."; - if (MONGO_FAIL_POINT(hangRefineCollectionShardKeyBeforeCommit)) { + if (MONGO_unlikely(hangRefineCollectionShardKeyBeforeCommit.shouldFail())) { log() << "Hit hangRefineCollectionShardKeyBeforeCommit failpoint"; - MONGO_FAIL_POINT_PAUSE_WHILE_SET_OR_INTERRUPTED( - opCtx, hangRefineCollectionShardKeyBeforeCommit); + hangRefineCollectionShardKeyBeforeCommit.pauseWhileSet(opCtx); } uassertStatusOK(commitTxnForConfigDocument(asr.opCtx(), txnNumber)); diff --git a/src/mongo/db/s/metadata_manager.cpp b/src/mongo/db/s/metadata_manager.cpp index e843eeff898..5d832418367 100644 --- a/src/mongo/db/s/metadata_manager.cpp +++ b/src/mongo/db/s/metadata_manager.cpp @@ -145,7 +145,7 @@ void scheduleCleanup(executor::TaskExecutor* executor, auto uniqueOpCtx = Client::getCurrent()->makeOperationContext(); auto opCtx = uniqueOpCtx.get(); - MONGO_FAIL_POINT_PAUSE_WHILE_SET(suspendRangeDeletion); + suspendRangeDeletion.pauseWhileSet(); auto next = CollectionRangeDeleter::cleanUpNextRange(opCtx, nss, epoch); if (next) { diff --git a/src/mongo/db/s/migration_destination_manager.cpp b/src/mongo/db/s/migration_destination_manager.cpp index e4b987f537c..51d6e4ba712 100644 --- a/src/mongo/db/s/migration_destination_manager.cpp +++ b/src/mongo/db/s/migration_destination_manager.cpp @@ -724,7 +724,7 @@ void MigrationDestinationManager::_migrateThread() { _setStateFail(str::stream() << "migrate failed: " << redact(exceptionToStatus())); } - if (getState() != DONE && !MONGO_FAIL_POINT(failMigrationLeaveOrphans)) { + if (getState() != DONE && !MONGO_unlikely(failMigrationLeaveOrphans.shouldFail())) { _forgetPending(opCtx.get(), ChunkRange(_min, _max)); } @@ -761,7 +761,7 @@ void MigrationDestinationManager::_migrateDriver(OperationContext* opCtx) { cloneCollectionIndexesAndOptions(opCtx, _nss, _fromShard); timing.done(1); - MONGO_FAIL_POINT_PAUSE_WHILE_SET(migrateThreadHangAtStep1); + migrateThreadHangAtStep1.pauseWhileSet(); } auto fromShard = @@ -787,7 +787,7 @@ void MigrationDestinationManager::_migrateDriver(OperationContext* opCtx) { } timing.done(2); - MONGO_FAIL_POINT_PAUSE_WHILE_SET(migrateThreadHangAtStep2); + migrateThreadHangAtStep2.pauseWhileSet(); } repl::OpTime lastOpApplied; @@ -884,9 +884,9 @@ void MigrationDestinationManager::_migrateDriver(OperationContext* opCtx) { lastOpApplied = cloneDocumentsFromDonor(opCtx, insertBatchFn, fetchBatchFn); timing.done(3); - MONGO_FAIL_POINT_PAUSE_WHILE_SET(migrateThreadHangAtStep3); + migrateThreadHangAtStep3.pauseWhileSet(); - if (MONGO_FAIL_POINT(failMigrationLeaveOrphans)) { + if (MONGO_unlikely(failMigrationLeaveOrphans.shouldFail())) { _setStateFail(str::stream() << "failing migration after cloning " << _numCloned << " docs due to failMigrationLeaveOrphans failpoint"); return; @@ -949,7 +949,7 @@ void MigrationDestinationManager::_migrateDriver(OperationContext* opCtx) { } timing.done(4); - MONGO_FAIL_POINT_PAUSE_WHILE_SET(migrateThreadHangAtStep4); + migrateThreadHangAtStep4.pauseWhileSet(); } { @@ -1024,7 +1024,7 @@ void MigrationDestinationManager::_migrateDriver(OperationContext* opCtx) { } timing.done(5); - MONGO_FAIL_POINT_PAUSE_WHILE_SET(migrateThreadHangAtStep5); + migrateThreadHangAtStep5.pauseWhileSet(); } _sessionMigration->join(); @@ -1036,7 +1036,7 @@ void MigrationDestinationManager::_migrateDriver(OperationContext* opCtx) { setState(DONE); timing.done(6); - MONGO_FAIL_POINT_PAUSE_WHILE_SET(migrateThreadHangAtStep6); + migrateThreadHangAtStep6.pauseWhileSet(); } bool MigrationDestinationManager::_applyMigrateOp(OperationContext* opCtx, @@ -1067,7 +1067,7 @@ bool MigrationDestinationManager::_applyMigrateOp(OperationContext* opCtx, BSONObj fullObj; if (Helpers::findById(opCtx, autoColl.getDb(), _nss.ns(), id, fullObj)) { if (!isInRange(fullObj, _min, _max, _shardKeyPattern)) { - if (MONGO_FAIL_POINT(failMigrationReceivedOutOfRangeOperation)) { + if (MONGO_unlikely(failMigrationReceivedOutOfRangeOperation.shouldFail())) { MONGO_UNREACHABLE; } continue; @@ -1105,7 +1105,7 @@ bool MigrationDestinationManager::_applyMigrateOp(OperationContext* opCtx, // do not apply insert/update if doc does not belong to the chunk being migrated if (!isInRange(updatedDoc, _min, _max, _shardKeyPattern)) { - if (MONGO_FAIL_POINT(failMigrationReceivedOutOfRangeOperation)) { + if (MONGO_unlikely(failMigrationReceivedOutOfRangeOperation.shouldFail())) { MONGO_UNREACHABLE; } continue; diff --git a/src/mongo/db/s/migration_source_manager.cpp b/src/mongo/db/s/migration_source_manager.cpp index 725af2a1ea7..56d84023c24 100644 --- a/src/mongo/db/s/migration_source_manager.cpp +++ b/src/mongo/db/s/migration_source_manager.cpp @@ -364,7 +364,7 @@ Status MigrationSourceManager::commitChunkOnRecipient(OperationContext* opCtx) { // Tell the recipient shard to fetch the latest changes. auto commitCloneStatus = _cloneDriver->commitClone(opCtx); - if (MONGO_FAIL_POINT(failMigrationCommit) && commitCloneStatus.isOK()) { + if (MONGO_unlikely(failMigrationCommit.shouldFail()) && commitCloneStatus.isOK()) { commitCloneStatus = {ErrorCodes::InternalError, "Failing _recvChunkCommit due to failpoint."}; } @@ -423,7 +423,7 @@ Status MigrationSourceManager::commitChunkMetadataOnConfig(OperationContext* opC builder.obj(), Shard::RetryPolicy::kIdempotent); - if (MONGO_FAIL_POINT(migrationCommitNetworkError)) { + if (MONGO_unlikely(migrationCommitNetworkError.shouldFail())) { commitChunkMigrationResponse = Status( ErrorCodes::InternalError, "Failpoint 'migrationCommitNetworkError' generated error"); } @@ -547,7 +547,7 @@ Status MigrationSourceManager::commitChunkMetadataOnConfig(OperationContext* opC LOG(0) << "Migration succeeded and updated collection version to " << refreshedMetadata->getCollVersion(); - MONGO_FAIL_POINT_PAUSE_WHILE_SET(hangBeforeLeavingCriticalSection); + hangBeforeLeavingCriticalSection.pauseWhileSet(); scopedGuard.dismiss(); @@ -576,7 +576,7 @@ Status MigrationSourceManager::commitChunkMetadataOnConfig(OperationContext* opC return CollectionShardingRuntime::get(opCtx, getNss())->cleanUpRange(range, whenToClean); }(); - if (!MONGO_FAIL_POINT(doNotRefreshRecipientAfterCommit)) { + if (!MONGO_unlikely(doNotRefreshRecipientAfterCommit.shouldFail())) { // Best-effort make the recipient refresh its routing table to the new collection version. refreshRecipientRoutingTable(opCtx, getNss(), diff --git a/src/mongo/db/s/move_chunk_command.cpp b/src/mongo/db/s/move_chunk_command.cpp index 8fafb8c0253..24c0023a6c4 100644 --- a/src/mongo/db/s/move_chunk_command.cpp +++ b/src/mongo/db/s/move_chunk_command.cpp @@ -217,30 +217,30 @@ private: moveChunkRequest.getFromShardId()); moveTimingHelper.done(1); - MONGO_FAIL_POINT_PAUSE_WHILE_SET(moveChunkHangAtStep1); + moveChunkHangAtStep1.pauseWhileSet(); MigrationSourceManager migrationSourceManager( opCtx, moveChunkRequest, donorConnStr, recipientHost); moveTimingHelper.done(2); - MONGO_FAIL_POINT_PAUSE_WHILE_SET(moveChunkHangAtStep2); + moveChunkHangAtStep2.pauseWhileSet(); uassertStatusOKWithWarning(migrationSourceManager.startClone(opCtx)); moveTimingHelper.done(3); - MONGO_FAIL_POINT_PAUSE_WHILE_SET(moveChunkHangAtStep3); + moveChunkHangAtStep3.pauseWhileSet(); uassertStatusOKWithWarning(migrationSourceManager.awaitToCatchUp(opCtx)); moveTimingHelper.done(4); - MONGO_FAIL_POINT_PAUSE_WHILE_SET(moveChunkHangAtStep4); + moveChunkHangAtStep4.pauseWhileSet(); uassertStatusOKWithWarning(migrationSourceManager.enterCriticalSection(opCtx)); uassertStatusOKWithWarning(migrationSourceManager.commitChunkOnRecipient(opCtx)); moveTimingHelper.done(5); - MONGO_FAIL_POINT_PAUSE_WHILE_SET(moveChunkHangAtStep5); + moveChunkHangAtStep5.pauseWhileSet(); uassertStatusOKWithWarning(migrationSourceManager.commitChunkMetadataOnConfig(opCtx)); moveTimingHelper.done(6); - MONGO_FAIL_POINT_PAUSE_WHILE_SET(moveChunkHangAtStep6); + moveChunkHangAtStep6.pauseWhileSet(); } } moveChunkCmd; diff --git a/src/mongo/db/s/shard_filtering_metadata_refresh.cpp b/src/mongo/db/s/shard_filtering_metadata_refresh.cpp index cab3a7391fd..9042fba9b5d 100644 --- a/src/mongo/db/s/shard_filtering_metadata_refresh.cpp +++ b/src/mongo/db/s/shard_filtering_metadata_refresh.cpp @@ -90,7 +90,7 @@ void onShardVersionMismatch(OperationContext* opCtx, } } - if (MONGO_FAIL_POINT(skipShardFilteringMetadataRefresh)) { + if (MONGO_unlikely(skipShardFilteringMetadataRefresh.shouldFail())) { return; } @@ -117,7 +117,7 @@ void onDbVersionMismatch(OperationContext* opCtx, // StaleDatabaseVersion retry attempts while the movePrimary is being committed. OperationShardingState::get(opCtx).waitForMovePrimaryCriticalSectionSignal(opCtx); - if (MONGO_FAIL_POINT(skipDatabaseVersionMetadataRefresh)) { + if (MONGO_unlikely(skipDatabaseVersionMetadataRefresh.shouldFail())) { return; } diff --git a/src/mongo/db/s/shard_server_catalog_cache_loader.cpp b/src/mongo/db/s/shard_server_catalog_cache_loader.cpp index cf158829ef5..78600e5d488 100644 --- a/src/mongo/db/s/shard_server_catalog_cache_loader.cpp +++ b/src/mongo/db/s/shard_server_catalog_cache_loader.cpp @@ -88,10 +88,9 @@ void dropChunksIfEpochChanged(OperationContext* opCtx, // table cache. dropChunks(opCtx, nss); - if (MONGO_FAIL_POINT(hangPersistCollectionAndChangedChunksAfterDropChunks)) { + if (MONGO_unlikely(hangPersistCollectionAndChangedChunksAfterDropChunks.shouldFail())) { log() << "Hit hangPersistCollectionAndChangedChunksAfterDropChunks failpoint"; - MONGO_FAIL_POINT_PAUSE_WHILE_SET_OR_INTERRUPTED( - opCtx, hangPersistCollectionAndChangedChunksAfterDropChunks); + hangPersistCollectionAndChangedChunksAfterDropChunks.pauseWhileSet(opCtx); } } } diff --git a/src/mongo/db/s/sharding_initialization_mongod.cpp b/src/mongo/db/s/sharding_initialization_mongod.cpp index 239f3c6bed2..29f9f699848 100644 --- a/src/mongo/db/s/sharding_initialization_mongod.cpp +++ b/src/mongo/db/s/sharding_initialization_mongod.cpp @@ -111,7 +111,7 @@ public: LOG(0) << "Updating config server with confirmed set " << connStr; Grid::get(serviceContext)->shardRegistry()->updateReplSetHosts(connStr); - if (MONGO_FAIL_POINT(failUpdateShardIdentityConfigString)) { + if (MONGO_unlikely(failUpdateShardIdentityConfigString.shouldFail())) { return; } diff --git a/src/mongo/db/s/shardsvr_rename_collection.cpp b/src/mongo/db/s/shardsvr_rename_collection.cpp index 84d4bda3c5f..41971a44c6e 100644 --- a/src/mongo/db/s/shardsvr_rename_collection.cpp +++ b/src/mongo/db/s/shardsvr_rename_collection.cpp @@ -72,10 +72,9 @@ public: uassertStatusOK(ActiveRenameCollectionRegistry::get(opCtx).registerRenameCollection( incomingRequest)); - if (MONGO_FAIL_POINT(hangRenameCollectionAfterGettingRename)) { + if (MONGO_unlikely(hangRenameCollectionAfterGettingRename.shouldFail())) { log() << "Hit hangRenameCollectionAfterGettingRename"; - MONGO_FAIL_POINT_PAUSE_WHILE_SET_OR_INTERRUPTED( - opCtx, hangRenameCollectionAfterGettingRename); + hangRenameCollectionAfterGettingRename.pauseWhileSet(opCtx); } // Check if there is an existing renameCollection running and if so, join it diff --git a/src/mongo/db/s/transaction_coordinator.cpp b/src/mongo/db/s/transaction_coordinator.cpp index c46920126ee..7836c63ef62 100644 --- a/src/mongo/db/s/transaction_coordinator.cpp +++ b/src/mongo/db/s/transaction_coordinator.cpp @@ -61,20 +61,19 @@ ExecutorFuture<void> waitForMajorityWithHangFailpoint(ServiceContext* service, return WaitForMajorityService::get(service).waitUntilMajority(opTime).thenRunOn(executor); }; - MONGO_FAIL_POINT_BLOCK(failpoint, fp) { + if (auto sfp = failpoint.scoped(); MONGO_unlikely(sfp.isActive())) { + const BSONObj& data = sfp.getData(); LOG(0) << "Hit " << failPointName << " failpoint"; - const BSONObj& data = fp.getData(); // Run the hang failpoint asynchronously on a different thread to avoid self deadlocks. return ExecutorFuture<void>(executor).then( [service, &failpoint, failPointName, data, waitForWC, opTime] { if (!data["useUninterruptibleSleep"].eoo()) { - MONGO_FAIL_POINT_PAUSE_WHILE_SET(failpoint); + failpoint.pauseWhileSet(); } else { ThreadClient tc(failPointName, service); auto opCtx = tc->makeOperationContext(); - - MONGO_FAIL_POINT_PAUSE_WHILE_SET_OR_INTERRUPTED(opCtx.get(), failpoint); + failpoint.pauseWhileSet(opCtx.get()); } return waitForWC(std::move(opTime)); diff --git a/src/mongo/db/s/transaction_coordinator_futures_util.cpp b/src/mongo/db/s/transaction_coordinator_futures_util.cpp index 59ae621bc59..5d3cf3bfdd5 100644 --- a/src/mongo/db/s/transaction_coordinator_futures_util.cpp +++ b/src/mongo/db/s/transaction_coordinator_futures_util.cpp @@ -95,9 +95,9 @@ Future<executor::TaskExecutor::ResponseStatus> AsyncWorkScheduler::scheduleRemot AuthorizationSession::get(opCtx->getClient()) ->grantInternalAuthorization(opCtx->getClient()); - if (MONGO_FAIL_POINT(hangWhileTargetingLocalHost)) { + if (MONGO_unlikely(hangWhileTargetingLocalHost.shouldFail())) { LOG(0) << "Hit hangWhileTargetingLocalHost failpoint"; - MONGO_FAIL_POINT_PAUSE_WHILE_SET_OR_INTERRUPTED(opCtx, hangWhileTargetingLocalHost); + hangWhileTargetingLocalHost.pauseWhileSet(opCtx); } const auto service = opCtx->getServiceContext(); @@ -231,9 +231,9 @@ Future<AsyncWorkScheduler::HostAndShard> AsyncWorkScheduler::_targetHostAsync( const auto shardRegistry = Grid::get(opCtx)->shardRegistry(); const auto shard = uassertStatusOK(shardRegistry->getShard(opCtx, shardId)); - if (MONGO_FAIL_POINT(hangWhileTargetingRemoteHost)) { + if (MONGO_unlikely(hangWhileTargetingRemoteHost.shouldFail())) { LOG(0) << "Hit hangWhileTargetingRemoteHost failpoint for shard " << shardId; - MONGO_FAIL_POINT_PAUSE_WHILE_SET_OR_INTERRUPTED(opCtx, hangWhileTargetingRemoteHost); + hangWhileTargetingRemoteHost.pauseWhileSet(opCtx); } // TODO (SERVER-35678): Return a SemiFuture<HostAndShard> rather than using a blocking call diff --git a/src/mongo/db/s/transaction_coordinator_util.cpp b/src/mongo/db/s/transaction_coordinator_util.cpp index 9e6b78c445a..c98a1c25775 100644 --- a/src/mongo/db/s/transaction_coordinator_util.cpp +++ b/src/mongo/db/s/transaction_coordinator_util.cpp @@ -107,9 +107,9 @@ repl::OpTime persistParticipantListBlocking(OperationContext* opCtx, const std::vector<ShardId>& participantList) { LOG(3) << txnIdToString(lsid, txnNumber) << " Going to write participant list"; - if (MONGO_FAIL_POINT(hangBeforeWritingParticipantList)) { + if (MONGO_unlikely(hangBeforeWritingParticipantList.shouldFail())) { LOG(0) << "Hit hangBeforeWritingParticipantList failpoint"; - MONGO_FAIL_POINT_PAUSE_WHILE_SET_OR_INTERRUPTED(opCtx, hangBeforeWritingParticipantList); + hangBeforeWritingParticipantList.pauseWhileSet(opCtx); } OperationSessionInfo sessionInfo; @@ -239,9 +239,9 @@ Future<PrepareVoteConsensus> sendPrepare(ServiceContext* service, getTransactionCoordinatorWorkerCurOpRepository()->set( opCtx, lsid, txnNumber, CoordinatorAction::kSendingPrepare); - if (MONGO_FAIL_POINT(hangBeforeSendingPrepare)) { + if (MONGO_unlikely(hangBeforeSendingPrepare.shouldFail())) { LOG(0) << "Hit hangBeforeSendingPrepare failpoint"; - MONGO_FAIL_POINT_PAUSE_WHILE_SET_OR_INTERRUPTED(opCtx, hangBeforeSendingPrepare); + hangBeforeSendingPrepare.pauseWhileSet(opCtx); } }; @@ -292,9 +292,9 @@ repl::OpTime persistDecisionBlocking(OperationContext* opCtx, LOG(3) << txnIdToString(lsid, txnNumber) << " Going to write decision " << (isCommit ? "commit" : "abort"); - if (MONGO_FAIL_POINT(hangBeforeWritingDecision)) { + if (MONGO_unlikely(hangBeforeWritingDecision.shouldFail())) { LOG(0) << "Hit hangBeforeWritingDecision failpoint"; - MONGO_FAIL_POINT_PAUSE_WHILE_SET_OR_INTERRUPTED(opCtx, hangBeforeWritingDecision); + hangBeforeWritingDecision.pauseWhileSet(opCtx); } OperationSessionInfo sessionInfo; @@ -401,9 +401,9 @@ Future<void> sendCommit(ServiceContext* service, getTransactionCoordinatorWorkerCurOpRepository()->set( opCtx, lsid, txnNumber, CoordinatorAction::kSendingCommit); - if (MONGO_FAIL_POINT(hangBeforeSendingCommit)) { + if (MONGO_unlikely(hangBeforeSendingCommit.shouldFail())) { LOG(0) << "Hit hangBeforeSendingCommit failpoint"; - MONGO_FAIL_POINT_PAUSE_WHILE_SET_OR_INTERRUPTED(opCtx, hangBeforeSendingCommit); + hangBeforeSendingCommit.pauseWhileSet(opCtx); } }; @@ -431,9 +431,9 @@ Future<void> sendAbort(ServiceContext* service, getTransactionCoordinatorWorkerCurOpRepository()->set( opCtx, lsid, txnNumber, CoordinatorAction::kSendingAbort); - if (MONGO_FAIL_POINT(hangBeforeSendingAbort)) { + if (MONGO_unlikely(hangBeforeSendingAbort.shouldFail())) { LOG(0) << "Hit hangBeforeSendingAbort failpoint"; - MONGO_FAIL_POINT_PAUSE_WHILE_SET_OR_INTERRUPTED(opCtx, hangBeforeSendingAbort); + hangBeforeSendingAbort.pauseWhileSet(opCtx); } }; @@ -451,9 +451,9 @@ void deleteCoordinatorDocBlocking(OperationContext* opCtx, TxnNumber txnNumber) { LOG(3) << txnIdToString(lsid, txnNumber) << " Going to delete coordinator doc"; - if (MONGO_FAIL_POINT(hangBeforeDeletingCoordinatorDoc)) { + if (MONGO_unlikely(hangBeforeDeletingCoordinatorDoc.shouldFail())) { LOG(0) << "Hit hangBeforeDeletingCoordinatorDoc failpoint"; - MONGO_FAIL_POINT_PAUSE_WHILE_SET_OR_INTERRUPTED(opCtx, hangBeforeDeletingCoordinatorDoc); + hangBeforeDeletingCoordinatorDoc.pauseWhileSet(opCtx); } OperationSessionInfo sessionInfo; @@ -506,15 +506,14 @@ void deleteCoordinatorDocBlocking(OperationContext* opCtx, LOG(3) << txnIdToString(lsid, txnNumber) << " Deleted coordinator doc"; - MONGO_FAIL_POINT_BLOCK(hangAfterDeletingCoordinatorDoc, fp) { + hangAfterDeletingCoordinatorDoc.execute([&](const BSONObj& data) { LOG(0) << "Hit hangAfterDeletingCoordinatorDoc failpoint"; - const BSONObj& data = fp.getData(); if (!data["useUninterruptibleSleep"].eoo()) { - MONGO_FAIL_POINT_PAUSE_WHILE_SET(hangAfterDeletingCoordinatorDoc); + hangAfterDeletingCoordinatorDoc.pauseWhileSet(); } else { - MONGO_FAIL_POINT_PAUSE_WHILE_SET_OR_INTERRUPTED(opCtx, hangAfterDeletingCoordinatorDoc); + hangAfterDeletingCoordinatorDoc.pauseWhileSet(opCtx); } - } + }); } } // namespace diff --git a/src/mongo/db/s/txn_two_phase_commit_cmds.cpp b/src/mongo/db/s/txn_two_phase_commit_cmds.cpp index 639b44e6e13..21293128b58 100644 --- a/src/mongo/db/s/txn_two_phase_commit_cmds.cpp +++ b/src/mongo/db/s/txn_two_phase_commit_cmds.cpp @@ -134,8 +134,8 @@ public: << opCtx->recoveryUnit()->getPrepareTimestamp().toString() << " participant prepareOpTime: " << prepareOpTime.toString()); - if (MONGO_FAIL_POINT( - participantReturnNetworkErrorForPrepareAfterExecutingPrepareLogic)) { + if (MONGO_unlikely(participantReturnNetworkErrorForPrepareAfterExecutingPrepareLogic + .shouldFail())) { uasserted(ErrorCodes::HostUnreachable, "returning network error because failpoint is on"); } @@ -143,8 +143,8 @@ public: } const auto prepareTimestamp = txnParticipant.prepareTransaction(opCtx, {}); - if (MONGO_FAIL_POINT( - participantReturnNetworkErrorForPrepareAfterExecutingPrepareLogic)) { + if (MONGO_unlikely(participantReturnNetworkErrorForPrepareAfterExecutingPrepareLogic + .shouldFail())) { uasserted(ErrorCodes::HostUnreachable, "returning network error because failpoint is on"); } @@ -234,10 +234,9 @@ public: *opCtx->getTxnNumber(), validateParticipants(opCtx, cmd.getParticipants())); - if (MONGO_FAIL_POINT(hangAfterStartingCoordinateCommit)) { + if (MONGO_unlikely(hangAfterStartingCoordinateCommit.shouldFail())) { LOG(0) << "Hit hangAfterStartingCoordinateCommit failpoint"; - MONGO_FAIL_POINT_PAUSE_WHILE_SET_OR_INTERRUPTED(opCtx, - hangAfterStartingCoordinateCommit); + hangAfterStartingCoordinateCommit.pauseWhileSet(opCtx); } ON_BLOCK_EXIT([opCtx] { diff --git a/src/mongo/db/server_options_helpers.cpp b/src/mongo/db/server_options_helpers.cpp index 9e06c0bc77c..a9d4badc2c6 100644 --- a/src/mongo/db/server_options_helpers.cpp +++ b/src/mongo/db/server_options_helpers.cpp @@ -146,7 +146,7 @@ Status validateBaseOptions(const moe::Environment& params) { if (enableTestCommandsValue) { // Only register failpoint server parameters if enableTestCommands=1. - getGlobalFailPointRegistry()->registerAllFailPointsAsServerParameters(); + globalFailPointRegistry().registerAllFailPointsAsServerParameters(); } else { // Deregister test-only parameters. ServerParameterSet::getGlobal()->disableTestParameters(); diff --git a/src/mongo/db/service_entry_point_common.cpp b/src/mongo/db/service_entry_point_common.cpp index f3760c3c8f4..b4b7d90ca59 100644 --- a/src/mongo/db/service_entry_point_common.cpp +++ b/src/mongo/db/service_entry_point_common.cpp @@ -574,16 +574,23 @@ bool runCommandImpl(OperationContext* opCtx, } auto waitForWriteConcern = [&](auto&& bb) { - MONGO_FAIL_POINT_BLOCK_IF(failCommand, data, [&](const BSONObj& data) { - return CommandHelpers::shouldActivateFailCommandFailPoint( - data, request.getCommandName(), opCtx->getClient(), invocation->ns()) && - data.hasField("writeConcernError"); - }) { - bb.append(data.getData()["writeConcernError"]); - return; // Don't do normal waiting. + bool reallyWait = true; + failCommand.executeIf( + [&](const BSONObj& data) { + bb.append(data["writeConcernError"]); + reallyWait = false; + }, + [&](const BSONObj& data) { + return CommandHelpers::shouldActivateFailCommandFailPoint( + data, + request.getCommandName(), + opCtx->getClient(), + invocation->ns()) && + data.hasField("writeConcernError"); + }); + if (reallyWait) { + behaviors.waitForWriteConcern(opCtx, invocation, lastOpBeforeRun, bb); } - - behaviors.waitForWriteConcern(opCtx, invocation, lastOpBeforeRun, bb); }; try { @@ -623,8 +630,7 @@ bool runCommandImpl(OperationContext* opCtx, // This failpoint should affect both getMores and commands which are read-only and thus don't // support writeConcern. if (!shouldWaitForWriteConcern || command->getLogicalOp() == LogicalOp::opGetMore) { - MONGO_FAIL_POINT_BLOCK(waitAfterReadCommandFinishesExecution, options) { - const BSONObj& data = options.getData(); + waitAfterReadCommandFinishesExecution.execute([&](const BSONObj& data) { auto db = data["db"].str(); if (db.empty() || request.getDatabase() == db) { CurOpFailpointHelpers::waitWhileFailPointEnabled( @@ -632,7 +638,7 @@ bool runCommandImpl(OperationContext* opCtx, opCtx, "waitAfterReadCommandFinishesExecution"); } - } + }); } behaviors.waitForLinearizableReadConcern(opCtx); @@ -692,15 +698,14 @@ void execCommandDatabase(OperationContext* opCtx, CurOp::get(opCtx)->setCommand_inlock(command); } - MONGO_FAIL_POINT_BLOCK(sleepMillisAfterCommandExecutionBegins, arg) { - const BSONObj& data = arg.getData(); + sleepMillisAfterCommandExecutionBegins.execute([&](const BSONObj& data) { auto numMillis = data["millis"].numberInt(); auto commands = data["commands"].Obj().getFieldNames<std::set<std::string>>(); // Only sleep for one of the specified commands. if (commands.find(command->getName()) != commands.end()) { mongo::sleepmillis(numMillis); } - } + }); // TODO: move this back to runCommands when mongos supports OperationContext // see SERVER-18515 for details. @@ -776,7 +781,7 @@ void execCommandDatabase(OperationContext* opCtx, const bool iAmPrimary = replCoord->canAcceptWritesForDatabase_UNSAFE(opCtx, dbname); if (!opCtx->getClient()->isInDirectClient() && - !MONGO_FAIL_POINT(skipCheckingForNotMasterInCommandDispatch)) { + !MONGO_unlikely(skipCheckingForNotMasterInCommandDispatch.shouldFail())) { const bool inMultiDocumentTransaction = (sessionOptions.getAutocommit() == false); auto allowed = command->secondaryAllowed(opCtx->getServiceContext()); bool alwaysAllowed = allowed == Command::AllowedOnSecondary::kAlways; @@ -789,7 +794,7 @@ void execCommandDatabase(OperationContext* opCtx, uasserted(ErrorCodes::NotMasterNoSlaveOk, "not master and slaveOk=false"); } - if (MONGO_FAIL_POINT(respondWithNotPrimaryInCommandDispatch)) { + if (MONGO_unlikely(respondWithNotPrimaryInCommandDispatch.shouldFail())) { uassert(ErrorCodes::NotMaster, "not primary", canRunHere); } else { uassert(ErrorCodes::NotMaster, "not master", canRunHere); @@ -1246,7 +1251,7 @@ DbResponse receivedGetMore(OperationContext* opCtx, audit::logGetMoreAuthzCheck(opCtx->getClient(), nsString, cursorid, status.code()); uassertStatusOK(status); - while (MONGO_FAIL_POINT(rsStopGetMore)) { + while (MONGO_unlikely(rsStopGetMore.shouldFail())) { sleepmillis(0); } diff --git a/src/mongo/db/service_entry_point_common.h b/src/mongo/db/service_entry_point_common.h index 87a6c178588..b41c0ef2c55 100644 --- a/src/mongo/db/service_entry_point_common.h +++ b/src/mongo/db/service_entry_point_common.h @@ -39,12 +39,12 @@ namespace mongo { -MONGO_FAIL_POINT_DECLARE(rsStopGetMore); -MONGO_FAIL_POINT_DECLARE(respondWithNotPrimaryInCommandDispatch); +extern FailPoint rsStopGetMore; +extern FailPoint respondWithNotPrimaryInCommandDispatch; // When active, we won't check if we are master in command dispatch. Activate this if you want to // test failing during command execution. -MONGO_FAIL_POINT_DECLARE(skipCheckingForNotMasterInCommandDispatch); +extern FailPoint skipCheckingForNotMasterInCommandDispatch; /** * Helpers for writing ServiceEntryPointImpl implementations from a reusable core. diff --git a/src/mongo/db/snapshot_window_util.cpp b/src/mongo/db/snapshot_window_util.cpp index 75d59c4abbd..05a46b42e13 100644 --- a/src/mongo/db/snapshot_window_util.cpp +++ b/src/mongo/db/snapshot_window_util.cpp @@ -88,7 +88,7 @@ void _decreaseTargetSnapshotWindowSize(WithLock lock, OperationContext* opCtx) { } // namespace void increaseTargetSnapshotWindowSize(OperationContext* opCtx) { - if (MONGO_FAIL_POINT(preventDynamicSnapshotHistoryWindowTargetAdjustments)) { + if (MONGO_unlikely(preventDynamicSnapshotHistoryWindowTargetAdjustments.shouldFail())) { return; } @@ -146,7 +146,7 @@ void increaseTargetSnapshotWindowSize(OperationContext* opCtx) { } void decreaseTargetSnapshotWindowSize(OperationContext* opCtx) { - if (MONGO_FAIL_POINT(preventDynamicSnapshotHistoryWindowTargetAdjustments)) { + if (MONGO_unlikely(preventDynamicSnapshotHistoryWindowTargetAdjustments.shouldFail())) { return; } diff --git a/src/mongo/db/storage/flow_control.cpp b/src/mongo/db/storage/flow_control.cpp index e811f87d4f5..ac010f891ae 100644 --- a/src/mongo/db/storage/flow_control.cpp +++ b/src/mongo/db/storage/flow_control.cpp @@ -294,8 +294,8 @@ int FlowControl::getNumTickets() { // Flow Control is only enabled on nodes that can accept writes. const bool canAcceptWrites = _replCoord->canAcceptNonLocalWrites(); - MONGO_FAIL_POINT_BLOCK(flowControlTicketOverride, failpointObj) { - int numTickets = failpointObj.getData().getIntField("numTickets"); + if (auto sfp = flowControlTicketOverride.scoped(); MONGO_unlikely(sfp.isActive())) { + int numTickets = sfp.getData().getIntField("numTickets"); if (numTickets > 0 && canAcceptWrites) { return numTickets; } diff --git a/src/mongo/db/storage/wiredtiger/wiredtiger_index.cpp b/src/mongo/db/storage/wiredtiger/wiredtiger_index.cpp index b911c67f590..f3a2c93e709 100644 --- a/src/mongo/db/storage/wiredtiger/wiredtiger_index.cpp +++ b/src/mongo/db/storage/wiredtiger/wiredtiger_index.cpp @@ -1125,7 +1125,7 @@ protected: std::memcmp(_key.getBuffer(), item.data, std::min(_key.getSize(), item.size)); bool nextNotIncreasing = cmp > 0 || (cmp == 0 && _key.getSize() > item.size); - if (MONGO_FAIL_POINT(WTEmulateOutOfOrderNextIndexKey)) { + if (MONGO_unlikely(WTEmulateOutOfOrderNextIndexKey.shouldFail())) { log() << "WTIndex::updatePosition simulating next key not increasing."; nextNotIncreasing = true; } diff --git a/src/mongo/db/storage/wiredtiger/wiredtiger_kv_engine.cpp b/src/mongo/db/storage/wiredtiger/wiredtiger_kv_engine.cpp index 1bad2e82872..556a01a2efb 100644 --- a/src/mongo/db/storage/wiredtiger/wiredtiger_kv_engine.cpp +++ b/src/mongo/db/storage/wiredtiger/wiredtiger_kv_engine.cpp @@ -1663,7 +1663,7 @@ void WiredTigerKVEngine::setOldestTimestampFromStable() { // Set the oldest timestamp to the stable timestamp to ensure that there is no lag window // between the two. - if (MONGO_FAIL_POINT(WTSetOldestTSToStableTS)) { + if (MONGO_unlikely(WTSetOldestTSToStableTS.shouldFail())) { setOldestTimestamp(stableTimestamp, false); return; } @@ -1681,7 +1681,7 @@ void WiredTigerKVEngine::setOldestTimestampFromStable() { } void WiredTigerKVEngine::setOldestTimestamp(Timestamp newOldestTimestamp, bool force) { - if (MONGO_FAIL_POINT(WTPreserveSnapshotHistoryIndefinitely)) { + if (MONGO_unlikely(WTPreserveSnapshotHistoryIndefinitely.shouldFail())) { return; } diff --git a/src/mongo/db/storage/wiredtiger/wiredtiger_oplog_manager.cpp b/src/mongo/db/storage/wiredtiger/wiredtiger_oplog_manager.cpp index 2edf06fb8a2..fbf0b9450a3 100644 --- a/src/mongo/db/storage/wiredtiger/wiredtiger_oplog_manager.cpp +++ b/src/mongo/db/storage/wiredtiger/wiredtiger_oplog_manager.cpp @@ -205,7 +205,7 @@ void WiredTigerOplogManager::_oplogJournalThreadLoop(WiredTigerSessionCache* ses } } - while (!_shuttingDown && MONGO_FAIL_POINT(WTPausePrimaryOplogDurabilityLoop)) { + while (!_shuttingDown && MONGO_unlikely(WTPausePrimaryOplogDurabilityLoop.shouldFail())) { lk.unlock(); sleepmillis(10); lk.lock(); diff --git a/src/mongo/db/storage/wiredtiger/wiredtiger_prepare_conflict.h b/src/mongo/db/storage/wiredtiger/wiredtiger_prepare_conflict.h index d2de945877d..8ce0a079bd4 100644 --- a/src/mongo/db/storage/wiredtiger/wiredtiger_prepare_conflict.h +++ b/src/mongo/db/storage/wiredtiger/wiredtiger_prepare_conflict.h @@ -38,12 +38,12 @@ namespace mongo { // When set, simulates returning WT_PREPARE_CONFLICT on WT cursor read operations. -MONGO_FAIL_POINT_DECLARE(WTPrepareConflictForReads); +extern FailPoint WTPrepareConflictForReads; // When set, WT_ROLLBACK is returned in place of retrying on WT_PREPARE_CONFLICT errors. -MONGO_FAIL_POINT_DECLARE(WTSkipPrepareConflictRetries); +extern FailPoint WTSkipPrepareConflictRetries; -MONGO_FAIL_POINT_DECLARE(WTPrintPrepareConflictLog); +extern FailPoint WTPrintPrepareConflictLog; /** * Logs a message with the number of prepare conflict retry attempts. @@ -72,8 +72,8 @@ int wiredTigerPrepareConflictRetry(OperationContext* opCtx, F&& f) { // error other than WT_PREPARE_CONFLICT. Reset PrepareConflictTracker accordingly. ON_BLOCK_EXIT([opCtx] { PrepareConflictTracker::get(opCtx).endPrepareConflict(); }); // If the failpoint is enabled, don't call the function, just simulate a conflict. - int ret = - MONGO_FAIL_POINT(WTPrepareConflictForReads) ? WT_PREPARE_CONFLICT : WT_READ_CHECK(f()); + int ret = MONGO_unlikely(WTPrepareConflictForReads.shouldFail()) ? WT_PREPARE_CONFLICT + : WT_READ_CHECK(f()); if (ret != WT_PREPARE_CONFLICT) return ret; @@ -85,7 +85,7 @@ int wiredTigerPrepareConflictRetry(OperationContext* opCtx, F&& f) { // this way are expected to be set to ignore prepare conflicts. invariant(!opCtx->isIgnoringInterrupts()); - if (MONGO_FAIL_POINT(WTPrintPrepareConflictLog)) { + if (MONGO_unlikely(WTPrintPrepareConflictLog.shouldFail())) { wiredTigerPrepareConflictFailPointLog(); } @@ -111,7 +111,7 @@ int wiredTigerPrepareConflictRetry(OperationContext* opCtx, F&& f) { str::stream() << lock.resourceId.toString() << " in " << modeName(lock.mode)); } - if (MONGO_FAIL_POINT(WTSkipPrepareConflictRetries)) { + if (MONGO_unlikely(WTSkipPrepareConflictRetries.shouldFail())) { // Callers of wiredTigerPrepareConflictRetry() should eventually call wtRCToStatus() via // invariantWTOK() and have the WT_ROLLBACK error bubble up as a WriteConflictException. // Enabling the "skipWriteConflictRetries" failpoint in conjunction with the @@ -124,8 +124,8 @@ int wiredTigerPrepareConflictRetry(OperationContext* opCtx, F&& f) { attempts++; auto lastCount = recoveryUnit->getSessionCache()->getPrepareCommitOrAbortCount(); // If the failpoint is enabled, don't call the function, just simulate a conflict. - ret = - MONGO_FAIL_POINT(WTPrepareConflictForReads) ? WT_PREPARE_CONFLICT : WT_READ_CHECK(f()); + ret = MONGO_unlikely(WTPrepareConflictForReads.shouldFail()) ? WT_PREPARE_CONFLICT + : WT_READ_CHECK(f()); if (ret != WT_PREPARE_CONFLICT) return ret; diff --git a/src/mongo/db/storage/wiredtiger/wiredtiger_record_store.h b/src/mongo/db/storage/wiredtiger/wiredtiger_record_store.h index e33e9a09f00..c0bf2784dc9 100644 --- a/src/mongo/db/storage/wiredtiger/wiredtiger_record_store.h +++ b/src/mongo/db/storage/wiredtiger/wiredtiger_record_store.h @@ -53,13 +53,14 @@ * conflict exception if the WTWriteConflictException failpoint is enabled. This is only checked * on cursor methods that make modifications. */ -#define WT_OP_CHECK(x) (((MONGO_FAIL_POINT(WTWriteConflictException))) ? (WT_ROLLBACK) : (x)) +#define WT_OP_CHECK(x) \ + (((MONGO_unlikely(WTWriteConflictException.shouldFail()))) ? (WT_ROLLBACK) : (x)) /** * Identical to WT_OP_CHECK except this is checked on cursor seeks/advancement. */ #define WT_READ_CHECK(x) \ - (((MONGO_FAIL_POINT(WTWriteConflictExceptionForReads))) ? (WT_ROLLBACK) : (x)) + (((MONGO_unlikely(WTWriteConflictExceptionForReads.shouldFail()))) ? (WT_ROLLBACK) : (x)) namespace mongo { @@ -520,11 +521,11 @@ private: // WT failpoint to throw write conflict exceptions randomly -MONGO_FAIL_POINT_DECLARE(WTWriteConflictException); -MONGO_FAIL_POINT_DECLARE(WTWriteConflictExceptionForReads); +extern FailPoint WTWriteConflictException; +extern FailPoint WTWriteConflictExceptionForReads; // Prevents oplog writes from being considered durable on the primary. Once activated, new writes // will not be considered durable until deactivated. It is unspecified whether writes that commit // before activation will become visible while active. -MONGO_FAIL_POINT_DECLARE(WTPausePrimaryOplogDurabilityLoop); +extern FailPoint WTPausePrimaryOplogDurabilityLoop; } // namespace mongo diff --git a/src/mongo/db/storage/wiredtiger/wiredtiger_recovery_unit.cpp b/src/mongo/db/storage/wiredtiger/wiredtiger_recovery_unit.cpp index b0275a2c7ab..e6abd29ed08 100644 --- a/src/mongo/db/storage/wiredtiger/wiredtiger_recovery_unit.cpp +++ b/src/mongo/db/storage/wiredtiger/wiredtiger_recovery_unit.cpp @@ -185,7 +185,7 @@ void WiredTigerRecoveryUnit::_commit() { } _setState(State::kCommitting); - if (MONGO_FAIL_POINT(WTAlwaysNotifyPrepareConflictWaiters)) { + if (MONGO_unlikely(WTAlwaysNotifyPrepareConflictWaiters.shouldFail())) { notifyDone = true; } @@ -204,7 +204,7 @@ void WiredTigerRecoveryUnit::_abort() { } _setState(State::kAborting); - if (notifyDone || MONGO_FAIL_POINT(WTAlwaysNotifyPrepareConflictWaiters)) { + if (notifyDone || MONGO_unlikely(WTAlwaysNotifyPrepareConflictWaiters.shouldFail())) { _sessionCache->notifyPreparedUnitOfWorkHasCommittedOrAborted(); } diff --git a/src/mongo/db/storage/write_unit_of_work.cpp b/src/mongo/db/storage/write_unit_of_work.cpp index c1dfa10d09a..8bddf4e9c59 100644 --- a/src/mongo/db/storage/write_unit_of_work.cpp +++ b/src/mongo/db/storage/write_unit_of_work.cpp @@ -102,7 +102,7 @@ void WriteUnitOfWork::commit() { invariant(!_released); invariant(_opCtx->_ruState == RecoveryUnitState::kActiveUnitOfWork); if (_toplevel) { - if (MONGO_FAIL_POINT(sleepBeforeCommit)) { + if (MONGO_unlikely(sleepBeforeCommit.shouldFail())) { sleepFor(Milliseconds(100)); } diff --git a/src/mongo/db/transaction_participant.cpp b/src/mongo/db/transaction_participant.cpp index e01ffa8a10e..26ab885031a 100644 --- a/src/mongo/db/transaction_participant.cpp +++ b/src/mongo/db/transaction_participant.cpp @@ -637,11 +637,11 @@ TransactionParticipant::OplogSlotReserver::OplogSlotReserver(OperationContext* o } TransactionParticipant::OplogSlotReserver::~OplogSlotReserver() { - if (MONGO_FAIL_POINT(hangBeforeReleasingTransactionOplogHole)) { + if (MONGO_unlikely(hangBeforeReleasingTransactionOplogHole.shouldFail())) { log() << "transaction - hangBeforeReleasingTransactionOplogHole fail point enabled. Blocking " "until fail point is disabled."; - MONGO_FAIL_POINT_PAUSE_WHILE_SET(hangBeforeReleasingTransactionOplogHole); + hangBeforeReleasingTransactionOplogHole.pauseWhileSet(); } // If the constructor did not complete, we do not attempt to abort the units of work. @@ -744,7 +744,7 @@ void TransactionParticipant::TxnResources::release(OperationContext* opCtx) { } _locker->reacquireTicket(opCtx); - if (MONGO_FAIL_POINT(restoreLocksFail)) { + if (MONGO_unlikely(restoreLocksFail.shouldFail())) { uasserted(ErrorCodes::LockTimeout, str::stream() << "Lock restore failed due to failpoint"); } @@ -1002,7 +1002,7 @@ void TransactionParticipant::Participant::unstashTransactionResources(OperationC // The Client lock must not be held when executing this failpoint as it will block currentOp // execution. - if (MONGO_FAIL_POINT(hangAfterPreallocateSnapshot)) { + if (MONGO_unlikely(hangAfterPreallocateSnapshot.shouldFail())) { CurOpFailpointHelpers::waitWhileFailPointEnabled( &hangAfterPreallocateSnapshot, opCtx, "hangAfterPreallocateSnapshot"); } @@ -1116,12 +1116,12 @@ Timestamp TransactionParticipant::Participant::prepareTransaction( o(lk).prepareOpTime = prepareOplogSlot; } - if (MONGO_FAIL_POINT(hangAfterReservingPrepareTimestamp)) { + if (MONGO_unlikely(hangAfterReservingPrepareTimestamp.shouldFail())) { // This log output is used in js tests so please leave it. log() << "transaction - hangAfterReservingPrepareTimestamp fail point " "enabled. Blocking until fail point is disabled. Prepare OpTime: " << prepareOplogSlot; - MONGO_FAIL_POINT_PAUSE_WHILE_SET(hangAfterReservingPrepareTimestamp); + hangAfterReservingPrepareTimestamp.pauseWhileSet(); } } opCtx->recoveryUnit()->setPrepareTimestamp(prepareOplogSlot.getTimestamp()); @@ -1145,10 +1145,10 @@ Timestamp TransactionParticipant::Participant::prepareTransaction( o(lk).lastWriteOpTime = prepareOplogSlot; } - if (MONGO_FAIL_POINT(hangAfterSettingPrepareStartTime)) { + if (MONGO_unlikely(hangAfterSettingPrepareStartTime.shouldFail())) { log() << "transaction - hangAfterSettingPrepareStartTime fail point enabled. Blocking " "until fail point is disabled."; - MONGO_FAIL_POINT_PAUSE_WHILE_SET(hangAfterSettingPrepareStartTime); + hangAfterSettingPrepareStartTime.pauseWhileSet(); } // We unlock the RSTL to allow prepared transactions to survive state transitions. This should @@ -1314,7 +1314,7 @@ void TransactionParticipant::Participant::commitPreparedTransaction( "commitTransaction for a prepared transaction cannot be run before its prepare " "oplog entry has been majority committed", replCoord->getLastCommittedOpTime().getTimestamp() >= prepareTimestamp || - MONGO_FAIL_POINT(skipCommitTxnCheckPrepareMajorityCommitted)); + MONGO_unlikely(skipCommitTxnCheckPrepareMajorityCommitted.shouldFail())); } try { @@ -2122,9 +2122,7 @@ void TransactionParticipant::Participant::_registerUpdateCacheOnCommit( } }); - MONGO_FAIL_POINT_BLOCK(onPrimaryTransactionalWrite, customArgs) { - const auto& data = customArgs.getData(); - + onPrimaryTransactionalWrite.execute([&](const BSONObj& data) { const auto closeConnectionElem = data["closeConnection"]; if (closeConnectionElem.eoo() || closeConnectionElem.Bool()) { opCtx->getClient()->session()->end(); @@ -2138,7 +2136,7 @@ void TransactionParticipant::Participant::_registerUpdateCacheOnCommit( << "Failing write for " << _sessionId() << ":" << o().activeTxnNumber << " due to failpoint. The write must not be reflected."); } - } + }); } } // namespace mongo diff --git a/src/mongo/db/transaction_participant_test.cpp b/src/mongo/db/transaction_participant_test.cpp index a638c8f6e2d..c5e58a96fbf 100644 --- a/src/mongo/db/transaction_participant_test.cpp +++ b/src/mongo/db/transaction_participant_test.cpp @@ -848,7 +848,7 @@ TEST_F(TxnParticipantTest, UnstashFailsShouldLeaveTxnResourceStashUnchanged) { ASSERT_FALSE(txnParticipant.getTxnResourceStashLockerForTest()->isLocked()); // Enable fail point. - getGlobalFailPointRegistry()->getFailPoint("restoreLocksFail")->setMode(FailPoint::alwaysOn); + globalFailPointRegistry().find("restoreLocksFail")->setMode(FailPoint::alwaysOn); ASSERT_THROWS_CODE(txnParticipant.unstashTransactionResources(opCtx(), "commitTransaction"), AssertionException, @@ -858,7 +858,7 @@ TEST_F(TxnParticipantTest, UnstashFailsShouldLeaveTxnResourceStashUnchanged) { ASSERT_FALSE(txnParticipant.getTxnResourceStashLockerForTest()->isLocked()); // Disable fail point. - getGlobalFailPointRegistry()->getFailPoint("restoreLocksFail")->setMode(FailPoint::off); + globalFailPointRegistry().find("restoreLocksFail")->setMode(FailPoint::off); // Should be successfully able to perform lock restore. txnParticipant.unstashTransactionResources(opCtx(), "commitTransaction"); diff --git a/src/mongo/db/ttl.cpp b/src/mongo/db/ttl.cpp index c7a5202b7b0..398ad6b1c8c 100644 --- a/src/mongo/db/ttl.cpp +++ b/src/mongo/db/ttl.cpp @@ -214,9 +214,9 @@ private: LOG(1) << "ns: " << collectionNSS << " key: " << key << " name: " << name; AutoGetCollection autoGetCollection(opCtx, collectionNSS, MODE_IX); - if (MONGO_FAIL_POINT(hangTTLMonitorWithLock)) { + if (MONGO_unlikely(hangTTLMonitorWithLock.shouldFail())) { log() << "Hanging due to hangTTLMonitorWithLock fail point"; - MONGO_FAIL_POINT_PAUSE_WHILE_SET_OR_INTERRUPTED(opCtx, hangTTLMonitorWithLock); + hangTTLMonitorWithLock.pauseWhileSet(opCtx); } diff --git a/src/mongo/db/write_concern.cpp b/src/mongo/db/write_concern.cpp index b4e2b756f9f..a1ea9c5bf92 100644 --- a/src/mongo/db/write_concern.cpp +++ b/src/mongo/db/write_concern.cpp @@ -171,7 +171,7 @@ Status waitForWriteConcern(OperationContext* opCtx, if (!opCtx->getClient()->isInDirectClient()) { // Respecting this failpoint for internal clients prevents stepup from working properly. - MONGO_FAIL_POINT_PAUSE_WHILE_SET(hangBeforeWaitingForWriteConcern); + hangBeforeWaitingForWriteConcern.pauseWhileSet(); } // Next handle blocking on disk diff --git a/src/mongo/executor/network_interface.h b/src/mongo/executor/network_interface.h index 6f938efb3a1..09a7fdaeb75 100644 --- a/src/mongo/executor/network_interface.h +++ b/src/mongo/executor/network_interface.h @@ -42,8 +42,8 @@ namespace mongo { namespace executor { -MONGO_FAIL_POINT_DECLARE(networkInterfaceDiscardCommandsBeforeAcquireConn); -MONGO_FAIL_POINT_DECLARE(networkInterfaceDiscardCommandsAfterAcquireConn); +extern FailPoint networkInterfaceDiscardCommandsBeforeAcquireConn; +extern FailPoint networkInterfaceDiscardCommandsAfterAcquireConn; /** * Interface to networking for use by TaskExecutor implementations. diff --git a/src/mongo/executor/network_interface_tl.cpp b/src/mongo/executor/network_interface_tl.cpp index 2cc0fc9244f..bcd1672e50e 100644 --- a/src/mongo/executor/network_interface_tl.cpp +++ b/src/mongo/executor/network_interface_tl.cpp @@ -286,7 +286,7 @@ Status NetworkInterfaceTL::startCommand(const TaskExecutor::CallbackHandle& cbHa } }); - if (MONGO_FAIL_POINT(networkInterfaceDiscardCommandsBeforeAcquireConn)) { + if (MONGO_unlikely(networkInterfaceDiscardCommandsBeforeAcquireConn.shouldFail())) { log() << "Discarding command due to failpoint before acquireConn"; return Status::OK(); } @@ -311,7 +311,7 @@ Status NetworkInterfaceTL::startCommand(const TaskExecutor::CallbackHandle& cbHa // We have a connection and the command hasn't already been attempted cmdState->request.emplace(cmdState->requestOnAny, idx); - if (MONGO_FAIL_POINT(networkInterfaceDiscardCommandsAfterAcquireConn)) { + if (MONGO_unlikely(networkInterfaceDiscardCommandsAfterAcquireConn.shouldFail())) { log() << "Discarding command due to failpoint after acquireConn"; return Status::OK(); } diff --git a/src/mongo/executor/scoped_task_executor.cpp b/src/mongo/executor/scoped_task_executor.cpp index a9b4e32dde8..e7e2f1bd5ae 100644 --- a/src/mongo/executor/scoped_task_executor.cpp +++ b/src/mongo/executor/scoped_task_executor.cpp @@ -215,10 +215,10 @@ private: std::piecewise_construct, std::forward_as_tuple(id), std::forward_as_tuple()); }; - if (MONGO_FAIL_POINT(ScopedTaskExecutorHangBeforeSchedule)) { + if (MONGO_unlikely(ScopedTaskExecutorHangBeforeSchedule.shouldFail())) { ScopedTaskExecutorHangBeforeSchedule.setMode(FailPoint::off); - MONGO_FAIL_POINT_PAUSE_WHILE_SET(ScopedTaskExecutorHangExitBeforeSchedule); + ScopedTaskExecutorHangExitBeforeSchedule.pauseWhileSet(); } // State 2 - Indeterminate state. We don't know yet if the task will get scheduled. @@ -259,7 +259,7 @@ private: doWorkAndNotify(args); }); - MONGO_FAIL_POINT_PAUSE_WHILE_SET(ScopedTaskExecutorHangAfterSchedule); + ScopedTaskExecutorHangAfterSchedule.pauseWhileSet(); stdx::unique_lock lk(_mutex); diff --git a/src/mongo/executor/scoped_task_executor.h b/src/mongo/executor/scoped_task_executor.h index 89d6187808f..dc166606115 100644 --- a/src/mongo/executor/scoped_task_executor.h +++ b/src/mongo/executor/scoped_task_executor.h @@ -107,9 +107,9 @@ private: std::shared_ptr<TaskExecutor> _executor; }; -MONGO_FAIL_POINT_DECLARE(ScopedTaskExecutorHangBeforeSchedule); -MONGO_FAIL_POINT_DECLARE(ScopedTaskExecutorHangExitBeforeSchedule); -MONGO_FAIL_POINT_DECLARE(ScopedTaskExecutorHangAfterSchedule); +extern FailPoint ScopedTaskExecutorHangBeforeSchedule; +extern FailPoint ScopedTaskExecutorHangExitBeforeSchedule; +extern FailPoint ScopedTaskExecutorHangAfterSchedule; } // namespace executor } // namespace mongo diff --git a/src/mongo/executor/scoped_task_executor_test.cpp b/src/mongo/executor/scoped_task_executor_test.cpp index ec078cb4dde..49c3842f392 100644 --- a/src/mongo/executor/scoped_task_executor_test.cpp +++ b/src/mongo/executor/scoped_task_executor_test.cpp @@ -288,7 +288,7 @@ TEST_F(ScopedTaskExecutorTest, scheduleLoseRaceWithShutdownOfUnderlying) { .isOK()); }); - MONGO_FAIL_POINT_PAUSE_WHILE_SET((bfp)); + (bfp).pauseWhileSet(); shutdownUnderlying(); diff --git a/src/mongo/executor/thread_pool_task_executor.cpp b/src/mongo/executor/thread_pool_task_executor.cpp index 808b2a7350c..55adfc29984 100644 --- a/src/mongo/executor/thread_pool_task_executor.cpp +++ b/src/mongo/executor/thread_pool_task_executor.cpp @@ -420,7 +420,7 @@ StatusWith<TaskExecutor::CallbackHandle> ThreadPoolTaskExecutor::scheduleRemoteC const RemoteCommandOnAnyCallbackFn& cb, const BatonHandle& baton) { - if (MONGO_FAIL_POINT(initialSyncFuzzerSynchronizationPoint1)) { + if (MONGO_unlikely(initialSyncFuzzerSynchronizationPoint1.shouldFail())) { // We are only going to pause on these failpoints if the command issued is for the // collection cloning part of initial sync. const auto cmdName = request.cmdObj.firstElementFieldName(); @@ -431,11 +431,11 @@ StatusWith<TaskExecutor::CallbackHandle> ThreadPoolTaskExecutor::scheduleRemoteC log() << "Collection Cloner scheduled a remote command on the " << request.dbname << " db: " << request.cmdObj; log() << "initialSyncFuzzerSynchronizationPoint1 fail point enabled."; - MONGO_FAIL_POINT_PAUSE_WHILE_SET(initialSyncFuzzerSynchronizationPoint1); + initialSyncFuzzerSynchronizationPoint1.pauseWhileSet(); - if (MONGO_FAIL_POINT(initialSyncFuzzerSynchronizationPoint2)) { + if (MONGO_unlikely(initialSyncFuzzerSynchronizationPoint2.shouldFail())) { log() << "initialSyncFuzzerSynchronizationPoint2 fail point enabled."; - MONGO_FAIL_POINT_PAUSE_WHILE_SET(initialSyncFuzzerSynchronizationPoint2); + initialSyncFuzzerSynchronizationPoint2.pauseWhileSet(); } } } @@ -600,7 +600,7 @@ void ThreadPoolTaskExecutor::scheduleIntoPool_inlock(WorkQueue* fromQueue, lk.unlock(); - if (MONGO_FAIL_POINT(scheduleIntoPoolSpinsUntilThreadPoolTaskExecutorShutsDown)) { + if (MONGO_unlikely(scheduleIntoPoolSpinsUntilThreadPoolTaskExecutorShutsDown.shouldFail())) { scheduleIntoPoolSpinsUntilThreadPoolTaskExecutorShutsDown.setMode(FailPoint::off); lk.lock(); diff --git a/src/mongo/executor/thread_pool_task_executor.h b/src/mongo/executor/thread_pool_task_executor.h index 218987383cb..9106f596069 100644 --- a/src/mongo/executor/thread_pool_task_executor.h +++ b/src/mongo/executor/thread_pool_task_executor.h @@ -44,8 +44,8 @@ namespace mongo { class ThreadPoolInterface; namespace executor { -MONGO_FAIL_POINT_DECLARE(initialSyncFuzzerSynchronizationPoint1); -MONGO_FAIL_POINT_DECLARE(initialSyncFuzzerSynchronizationPoint2); +extern FailPoint initialSyncFuzzerSynchronizationPoint1; +extern FailPoint initialSyncFuzzerSynchronizationPoint2; struct ConnectionPoolStats; class NetworkInterface; diff --git a/src/mongo/executor/thread_pool_task_executor_test.cpp b/src/mongo/executor/thread_pool_task_executor_test.cpp index ed458e50a7c..fb7330ca3ee 100644 --- a/src/mongo/executor/thread_pool_task_executor_test.cpp +++ b/src/mongo/executor/thread_pool_task_executor_test.cpp @@ -215,11 +215,11 @@ TEST_F(ThreadPoolExecutorTest, ShutdownAndScheduleWorkRaceDoesNotCrash) { }) .getStatus()); - auto fpTPTE1 = getGlobalFailPointRegistry()->getFailPoint( - "scheduleIntoPoolSpinsUntilThreadPoolTaskExecutorShutsDown"); + auto fpTPTE1 = + globalFailPointRegistry().find("scheduleIntoPoolSpinsUntilThreadPoolTaskExecutorShutsDown"); fpTPTE1->setMode(FailPoint::alwaysOn); barrier.countDownAndWait(); - MONGO_FAIL_POINT_PAUSE_WHILE_SET((*fpTPTE1)); + (*fpTPTE1).pauseWhileSet(); executor.shutdown(); executor.join(); ASSERT_OK(status1); @@ -247,11 +247,11 @@ TEST_F(ThreadPoolExecutorTest, ShutdownAndScheduleRaceDoesNotCrash) { amRunningRecursively = false; }); - auto fpTPTE1 = getGlobalFailPointRegistry()->getFailPoint( - "scheduleIntoPoolSpinsUntilThreadPoolTaskExecutorShutsDown"); + auto fpTPTE1 = + globalFailPointRegistry().find("scheduleIntoPoolSpinsUntilThreadPoolTaskExecutorShutsDown"); fpTPTE1->setMode(FailPoint::alwaysOn); barrier.countDownAndWait(); - MONGO_FAIL_POINT_PAUSE_WHILE_SET((*fpTPTE1)); + (*fpTPTE1).pauseWhileSet(); executor.shutdown(); executor.join(); ASSERT_OK(status1); diff --git a/src/mongo/rpc/op_msg_integration_test.cpp b/src/mongo/rpc/op_msg_integration_test.cpp index 8ee613a3616..5cd98239b5f 100644 --- a/src/mongo/rpc/op_msg_integration_test.cpp +++ b/src/mongo/rpc/op_msg_integration_test.cpp @@ -279,12 +279,12 @@ TEST(OpMsg, DocumentSequenceReturnsWork) { constexpr auto kDisableChecksum = "dbClientConnectionDisableChecksum"; void disableClientChecksum() { - auto failPoint = getGlobalFailPointRegistry()->getFailPoint(kDisableChecksum); + auto failPoint = globalFailPointRegistry().find(kDisableChecksum); failPoint->setMode(FailPoint::alwaysOn); } void enableClientChecksum() { - auto failPoint = getGlobalFailPointRegistry()->getFailPoint(kDisableChecksum); + auto failPoint = globalFailPointRegistry().find(kDisableChecksum); failPoint->setMode(FailPoint::off); } diff --git a/src/mongo/s/catalog/replset_dist_lock_manager.cpp b/src/mongo/s/catalog/replset_dist_lock_manager.cpp index 9a2eb86e060..860b464939d 100644 --- a/src/mongo/s/catalog/replset_dist_lock_manager.cpp +++ b/src/mongo/s/catalog/replset_dist_lock_manager.cpp @@ -304,10 +304,9 @@ StatusWith<DistLockHandle> ReplSetDistLockManager::lockWithSessionID(OperationCo const string who = str::stream() << _processID << ":" << getThreadName(); auto lockExpiration = _lockExpiration; - MONGO_FAIL_POINT_BLOCK(setDistLockTimeout, customTimeout) { - const BSONObj& data = customTimeout.getData(); + setDistLockTimeout.execute([&](const BSONObj& data) { lockExpiration = Milliseconds(data["timeoutMs"].numberInt()); - } + }); LOG(1) << "trying to acquire new distributed lock for " << name << " ( lock timeout : " << durationCount<Milliseconds>(lockExpiration) diff --git a/src/mongo/s/catalog/sharding_catalog_client_impl.cpp b/src/mongo/s/catalog/sharding_catalog_client_impl.cpp index 501f24f39a0..a73251cd98c 100644 --- a/src/mongo/s/catalog/sharding_catalog_client_impl.cpp +++ b/src/mongo/s/catalog/sharding_catalog_client_impl.cpp @@ -709,7 +709,7 @@ Status ShardingCatalogClientImpl::applyChunkOpsDeprecated(OperationContext* opCt // TODO (Dianna) This fail point needs to be reexamined when CommitChunkMigration is in: // migrations will no longer be able to exercise it, so split or merge will need to do so. // SERVER-22659. - if (MONGO_FAIL_POINT(failApplyChunkOps)) { + if (MONGO_unlikely(failApplyChunkOps.shouldFail())) { status = Status(ErrorCodes::InternalError, "Failpoint 'failApplyChunkOps' generated error"); } diff --git a/src/mongo/s/commands/cluster_refine_collection_shard_key_cmd.cpp b/src/mongo/s/commands/cluster_refine_collection_shard_key_cmd.cpp index 23d8d1a3314..ede2821a364 100644 --- a/src/mongo/s/commands/cluster_refine_collection_shard_key_cmd.cpp +++ b/src/mongo/s/commands/cluster_refine_collection_shard_key_cmd.cpp @@ -58,10 +58,9 @@ public: Grid::get(opCtx)->catalogCache()->getShardedCollectionRoutingInfoWithRefresh(opCtx, nss)); - if (MONGO_FAIL_POINT(hangRefineCollectionShardKeyAfterRefresh)) { + if (MONGO_unlikely(hangRefineCollectionShardKeyAfterRefresh.shouldFail())) { log() << "Hit hangRefineCollectionShardKeyAfterRefresh failpoint"; - MONGO_FAIL_POINT_PAUSE_WHILE_SET_OR_INTERRUPTED( - opCtx, hangRefineCollectionShardKeyAfterRefresh); + hangRefineCollectionShardKeyAfterRefresh.pauseWhileSet(opCtx); } ConfigsvrRefineCollectionShardKey configsvrRefineCollShardKey( diff --git a/src/mongo/s/commands/cluster_write_cmd.cpp b/src/mongo/s/commands/cluster_write_cmd.cpp index d7b10d8f88d..bc48977219d 100644 --- a/src/mongo/s/commands/cluster_write_cmd.cpp +++ b/src/mongo/s/commands/cluster_write_cmd.cpp @@ -204,10 +204,9 @@ bool handleWouldChangeOwningShardError(OperationContext* opCtx, bool updatedShardKey = false; boost::optional<BSONObj> upsertedId; if (isRetryableWrite) { - if (MONGO_FAIL_POINT(hangAfterThrowWouldChangeOwningShardRetryableWrite)) { + if (MONGO_unlikely(hangAfterThrowWouldChangeOwningShardRetryableWrite.shouldFail())) { log() << "Hit hangAfterThrowWouldChangeOwningShardRetryableWrite failpoint"; - MONGO_FAIL_POINT_PAUSE_WHILE_SET_OR_INTERRUPTED( - opCtx, hangAfterThrowWouldChangeOwningShardRetryableWrite); + hangAfterThrowWouldChangeOwningShardRetryableWrite.pauseWhileSet(opCtx); } RouterOperationContextSession routerSession(opCtx); try { diff --git a/src/mongo/s/commands/commands_public.cpp b/src/mongo/s/commands/commands_public.cpp index e038044ac49..20782145cfa 100644 --- a/src/mongo/s/commands/commands_public.cpp +++ b/src/mongo/s/commands/commands_public.cpp @@ -214,7 +214,7 @@ public: str::stream() << "Invalid target namespace: " << toNss.ns(), toNss.isValid()); - if (MONGO_FAIL_POINT(useRenameCollectionPathThroughConfigsvr)) { + if (MONGO_unlikely(useRenameCollectionPathThroughConfigsvr.shouldFail())) { bool dropTarget = cmdObj["dropTarget"].trueValue(); bool stayTemp = cmdObj["stayTemp"].trueValue(); diff --git a/src/mongo/s/commands/document_shard_key_update_util.cpp b/src/mongo/s/commands/document_shard_key_update_util.cpp index 119193d9ee5..90f12f4c8f5 100644 --- a/src/mongo/s/commands/document_shard_key_update_util.cpp +++ b/src/mongo/s/commands/document_shard_key_update_util.cpp @@ -77,9 +77,9 @@ bool executeOperationsAsPartOfShardKeyUpdate(OperationContext* opCtx, return false; } - if (MONGO_FAIL_POINT(hangBeforeInsertOnUpdateShardKey)) { + if (MONGO_unlikely(hangBeforeInsertOnUpdateShardKey.shouldFail())) { log() << "Hit hangBeforeInsertOnUpdateShardKey failpoint"; - MONGO_FAIL_POINT_PAUSE_WHILE_SET_OR_INTERRUPTED(opCtx, hangBeforeInsertOnUpdateShardKey); + hangBeforeInsertOnUpdateShardKey.pauseWhileSet(opCtx); } auto insertOpMsg = OpMsgRequest::fromDBAndBody(db, insertCmdObj); diff --git a/src/mongo/s/commands/strategy.cpp b/src/mongo/s/commands/strategy.cpp index cc93d78a591..5949128fe88 100644 --- a/src/mongo/s/commands/strategy.cpp +++ b/src/mongo/s/commands/strategy.cpp @@ -298,15 +298,15 @@ void execCommandClient(OperationContext* opCtx, invocation->run(opCtx, result); } - auto body = result->getBodyBuilder(); - - MONGO_FAIL_POINT_BLOCK_IF(failCommand, data, [&](const BSONObj& data) { - return CommandHelpers::shouldActivateFailCommandFailPoint( - data, request.getCommandName(), opCtx->getClient(), invocation->ns()) && - data.hasField("writeConcernError"); - }) { - body.append(data.getData()["writeConcernError"]); - } + failCommand.executeIf( + [&](const BSONObj& data) { + result->getBodyBuilder().append(data["writeConcernError"]); + }, + [&](const BSONObj& data) { + return CommandHelpers::shouldActivateFailCommandFailPoint( + data, request.getCommandName(), opCtx->getClient(), invocation->ns()) && + data.hasField("writeConcernError"); + }); } auto body = result->getBodyBuilder(); @@ -484,7 +484,7 @@ void runCommand(OperationContext* opCtx, // under a transaction (see the invariant inside ShardConnection). Because of this, // the retargeting error could not have come from a ShardConnection, so we don't // need to reset the connection's in-memory state. - if (!MONGO_FAIL_POINT(doNotRefreshShardsOnRetargettingError) && + if (!MONGO_unlikely(doNotRefreshShardsOnRetargettingError.shouldFail()) && !TransactionRouter::get(opCtx)) { ShardConnection::checkMyConnectionVersions(opCtx, staleNs.ns()); } @@ -990,7 +990,7 @@ void Strategy::explainFind(OperationContext* opCtx, // under a transaction (see the invariant inside ShardConnection). Because of this, the // retargeting error could not have come from a ShardConnection, so we don't need to // reset the connection's in-memory state. - if (!MONGO_FAIL_POINT(doNotRefreshShardsOnRetargettingError) && + if (!MONGO_unlikely(doNotRefreshShardsOnRetargettingError.shouldFail()) && !TransactionRouter::get(opCtx)) { ShardConnection::checkMyConnectionVersions(opCtx, staleNs.ns()); } diff --git a/src/mongo/s/query/cluster_aggregate.cpp b/src/mongo/s/query/cluster_aggregate.cpp index cfc60541fb3..953f4c48d05 100644 --- a/src/mongo/s/query/cluster_aggregate.cpp +++ b/src/mongo/s/query/cluster_aggregate.cpp @@ -147,7 +147,7 @@ sharded_agg_helpers::DispatchShardPipelineResults dispatchExchangeConsumerPipeli invariant(!litePipe.hasChangeStream()); auto opCtx = expCtx->opCtx; - if (MONGO_FAIL_POINT(clusterAggregateFailToDispatchExchangeConsumerPipeline)) { + if (MONGO_unlikely(clusterAggregateFailToDispatchExchangeConsumerPipeline.shouldFail())) { log() << "clusterAggregateFailToDispatchExchangeConsumerPipeline fail point enabled."; uasserted(ErrorCodes::FailPointEnabled, "Asserting on exhange consumer pipeline dispatch due to failpoint."); @@ -300,7 +300,7 @@ AsyncRequestsSender::Response establishMergingShardCursor(OperationContext* opCt const AggregationRequest& request, const BSONObj mergeCmdObj, const ShardId& mergingShardId) { - if (MONGO_FAIL_POINT(clusterAggregateFailToEstablishMergingShardCursor)) { + if (MONGO_unlikely(clusterAggregateFailToEstablishMergingShardCursor.shouldFail())) { log() << "clusterAggregateFailToEstablishMergingShardCursor fail point enabled."; uasserted(ErrorCodes::FailPointEnabled, "Asserting on establishing merging shard cursor due to failpoint."); diff --git a/src/mongo/s/query/cluster_find.cpp b/src/mongo/s/query/cluster_find.cpp index 274931769f9..a51435e0db4 100644 --- a/src/mongo/s/query/cluster_find.cpp +++ b/src/mongo/s/query/cluster_find.cpp @@ -601,7 +601,7 @@ StatusWith<CursorResponse> ClusterFind::runGetMore(OperationContext* opCtx, // If the 'waitAfterPinningCursorBeforeGetMoreBatch' fail point is enabled, set the 'msg' // field of this operation's CurOp to signal that we've hit this point. - if (MONGO_FAIL_POINT(waitAfterPinningCursorBeforeGetMoreBatch)) { + if (MONGO_unlikely(waitAfterPinningCursorBeforeGetMoreBatch.shouldFail())) { CurOpFailpointHelpers::waitWhileFailPointEnabled( &waitAfterPinningCursorBeforeGetMoreBatch, opCtx, @@ -624,7 +624,7 @@ StatusWith<CursorResponse> ClusterFind::runGetMore(OperationContext* opCtx, // If the 'waitWithPinnedCursorDuringGetMoreBatch' fail point is enabled, set the 'msg' // field of this operation's CurOp to signal that we've hit this point. - if (MONGO_FAIL_POINT(waitWithPinnedCursorDuringGetMoreBatch)) { + if (MONGO_unlikely(waitWithPinnedCursorDuringGetMoreBatch.shouldFail())) { CurOpFailpointHelpers::waitWhileFailPointEnabled(&waitWithPinnedCursorDuringGetMoreBatch, opCtx, "waitWithPinnedCursorDuringGetMoreBatch"); @@ -701,7 +701,7 @@ StatusWith<CursorResponse> ClusterFind::runGetMore(OperationContext* opCtx, CurOp::get(opCtx)->debug().cursorExhausted = (idToReturn == 0); CurOp::get(opCtx)->debug().nreturned = batch.size(); - if (MONGO_FAIL_POINT(waitBeforeUnpinningOrDeletingCursorAfterGetMoreBatch)) { + if (MONGO_unlikely(waitBeforeUnpinningOrDeletingCursorAfterGetMoreBatch.shouldFail())) { CurOpFailpointHelpers::waitWhileFailPointEnabled( &waitBeforeUnpinningOrDeletingCursorAfterGetMoreBatch, opCtx, diff --git a/src/mongo/s/server.cpp b/src/mongo/s/server.cpp index e39a8b00142..dc66e21307b 100644 --- a/src/mongo/s/server.cpp +++ b/src/mongo/s/server.cpp @@ -459,7 +459,7 @@ public: Grid::get(serviceContext)->shardRegistry()->updateReplSetHosts(connStr); - if (MONGO_FAIL_POINT(failReplicaSetChangeConfigServerUpdateHook)) { + if (MONGO_unlikely(failReplicaSetChangeConfigServerUpdateHook.shouldFail())) { return; } ShardRegistry::updateReplicaSetOnConfigServer(serviceContext, connStr); diff --git a/src/mongo/s/transaction_router.cpp b/src/mongo/s/transaction_router.cpp index 81cd77fcab0..b9a77045d82 100644 --- a/src/mongo/s/transaction_router.cpp +++ b/src/mongo/s/transaction_router.cpp @@ -719,7 +719,7 @@ void TransactionRouter::Router::_clearPendingParticipants(OperationContext* opCt } bool TransactionRouter::Router::canContinueOnStaleShardOrDbError(StringData cmdName) const { - if (MONGO_FAIL_POINT(enableStaleVersionAndSnapshotRetriesWithinTransactions)) { + if (MONGO_unlikely(enableStaleVersionAndSnapshotRetriesWithinTransactions.shouldFail())) { // We can always retry on the first overall statement because all targeted participants must // be pending, so the retry will restart the local transaction on each one, overwriting any // effects from the first attempt. @@ -770,7 +770,7 @@ void TransactionRouter::Router::onViewResolutionError(OperationContext* opCtx, } bool TransactionRouter::Router::canContinueOnSnapshotError() const { - if (MONGO_FAIL_POINT(enableStaleVersionAndSnapshotRetriesWithinTransactions)) { + if (MONGO_unlikely(enableStaleVersionAndSnapshotRetriesWithinTransactions.shouldFail())) { return o().atClusterTime && o().atClusterTime->canChange(p().latestStmtId); } diff --git a/src/mongo/scripting/engine.cpp b/src/mongo/scripting/engine.cpp index 0c477a24c11..b5093581889 100644 --- a/src/mongo/scripting/engine.cpp +++ b/src/mongo/scripting/engine.cpp @@ -237,7 +237,7 @@ void Scope::loadStored(OperationContext* opCtx, bool ignoreNotConnected) { uassert(10209, str::stream() << "name has to be a string: " << n, n.type() == String); uassert(10210, "value has to be set", v.type() != EOO); - if (MONGO_FAIL_POINT(mr_killop_test_fp)) { + if (MONGO_unlikely(mr_killop_test_fp.shouldFail())) { /* This thread sleep makes the interrupts in the test come in at a time * where the js misses the interrupt and throw an exception instead of diff --git a/src/mongo/transport/session_asio.h b/src/mongo/transport/session_asio.h index 484b11f3331..434371863e5 100644 --- a/src/mongo/transport/session_asio.h +++ b/src/mongo/transport/session_asio.h @@ -446,7 +446,7 @@ private: std::error_code ec; size_t size; - if (MONGO_FAIL_POINT(transportLayerASIOshortOpportunisticReadWrite) && + if (MONGO_unlikely(transportLayerASIOshortOpportunisticReadWrite.shouldFail()) && _blockingMode == Async) { asio::mutable_buffer localBuffer = buffers; @@ -529,7 +529,7 @@ private: std::error_code ec; std::size_t size; - if (MONGO_FAIL_POINT(transportLayerASIOshortOpportunisticReadWrite) && + if (MONGO_unlikely(transportLayerASIOshortOpportunisticReadWrite.shouldFail()) && _blockingMode == Async) { asio::const_buffer localBuffer = buffers; diff --git a/src/mongo/transport/transport_layer_asio.cpp b/src/mongo/transport/transport_layer_asio.cpp index cad586f38d6..58153a870e4 100644 --- a/src/mongo/transport/transport_layer_asio.cpp +++ b/src/mongo/transport/transport_layer_asio.cpp @@ -621,7 +621,7 @@ Future<SessionHandle> TransportLayerASIO::asyncConnect(HostAndPort peer, return makeConnectError(status, connector->peer, connector->resolvedEndpoint); }) .getAsync([connector](Status connectResult) { - if (MONGO_FAIL_POINT(transportLayerASIOasyncConnectTimesOut)) { + if (MONGO_unlikely(transportLayerASIOasyncConnectTimesOut.shouldFail())) { log() << "asyncConnectTimesOut fail point is active. simulating timeout."; return; } diff --git a/src/mongo/transport/transport_layer_asio.h b/src/mongo/transport/transport_layer_asio.h index c0e14452adf..90008fe3c89 100644 --- a/src/mongo/transport/transport_layer_asio.h +++ b/src/mongo/transport/transport_layer_asio.h @@ -69,11 +69,11 @@ class ServiceEntryPoint; namespace transport { // This fail point simulates reads and writes that always return 1 byte and fail with EAGAIN -MONGO_FAIL_POINT_DECLARE(transportLayerASIOshortOpportunisticReadWrite); +extern FailPoint transportLayerASIOshortOpportunisticReadWrite; // This fail point will cause an asyncConnect to timeout after it's successfully connected // to the remote peer -MONGO_FAIL_POINT_DECLARE(transportLayerASIOasyncConnectTimesOut); +extern FailPoint transportLayerASIOasyncConnectTimesOut; /** * A TransportLayer implementation based on ASIO networking primitives. diff --git a/src/mongo/util/diagnostic_info.cpp b/src/mongo/util/diagnostic_info.cpp index a42cc5275a9..a05dcb8a3f6 100644 --- a/src/mongo/util/diagnostic_info.cpp +++ b/src/mongo/util/diagnostic_info.cpp @@ -73,7 +73,7 @@ MONGO_INITIALIZER(LockActions)(InitializerContext* context) { DiagnosticInfo::Diagnostic::clearDiagnostic(); } void onFailedLock() override { - if (!MONGO_FAIL_POINT(keepDiagnosticCaptureOnFailedLock)) { + if (!MONGO_unlikely(keepDiagnosticCaptureOnFailedLock.shouldFail())) { DiagnosticInfo::Diagnostic::clearDiagnostic(); } } diff --git a/src/mongo/util/diagnostic_info.h b/src/mongo/util/diagnostic_info.h index 7f7325c9b23..0a4958f35de 100644 --- a/src/mongo/util/diagnostic_info.h +++ b/src/mongo/util/diagnostic_info.h @@ -36,7 +36,7 @@ #include "mongo/util/time_support.h" namespace mongo { -MONGO_FAIL_POINT_DECLARE(keepDiagnosticCaptureOnFailedLock); +extern FailPoint keepDiagnosticCaptureOnFailedLock; /** * DiagnosticInfo keeps track of diagnostic information such as a developer provided * name, the time when a lock was first acquired, and a partial caller call stack. diff --git a/src/mongo/util/fail_point.cpp b/src/mongo/util/fail_point.cpp index d5af18f318f..c40a4b829d6 100644 --- a/src/mongo/util/fail_point.cpp +++ b/src/mongo/util/fail_point.cpp @@ -84,7 +84,7 @@ void FailPoint::shouldFailCloseBlock() { _fpInfo.subtractAndFetch(1); } -void FailPoint::setMode(Mode mode, ValType val, const BSONObj& extra) { +void FailPoint::setMode(Mode mode, ValType val, BSONObj extra) { /** * Outline: * @@ -96,20 +96,20 @@ void FailPoint::setMode(Mode mode, ValType val, const BSONObj& extra) { stdx::lock_guard<stdx::mutex> scoped(_modMutex); // Step 1 - disableFailPoint(); + disable(); // Step 2 while (_fpInfo.load() != 0) { sleepmillis(50); } + // Step 3 _mode = mode; _timesOrPeriod.store(val); - - _data = extra.copy(); + _data = std::move(extra); if (_mode != off) { - enableFailPoint(); + enable(); } } @@ -117,19 +117,19 @@ const BSONObj& FailPoint::getData() const { return _data; } -void FailPoint::enableFailPoint() { - _fpInfo.fetchAndBitOr(ACTIVE_BIT); +void FailPoint::enable() { + _fpInfo.fetchAndBitOr(kActiveBit); } -void FailPoint::disableFailPoint() { - _fpInfo.fetchAndBitAnd(~ACTIVE_BIT); +void FailPoint::disable() { + _fpInfo.fetchAndBitAnd(~kActiveBit); } FailPoint::RetCode FailPoint::slowShouldFailOpenBlock( std::function<bool(const BSONObj&)> cb) noexcept { ValType localFpInfo = _fpInfo.addAndFetch(1); - if ((localFpInfo & ACTIVE_BIT) == 0) { + if ((localFpInfo & kActiveBit) == 0) { return slowOff; } @@ -149,7 +149,7 @@ FailPoint::RetCode FailPoint::slowShouldFailOpenBlock( } case nTimes: { if (_timesOrPeriod.subtractAndFetch(1) <= 0) - disableFailPoint(); + disable(); return slowOn; } @@ -168,8 +168,7 @@ FailPoint::RetCode FailPoint::slowShouldFailOpenBlock( } } -StatusWith<std::tuple<FailPoint::Mode, FailPoint::ValType, BSONObj>> FailPoint::parseBSON( - const BSONObj& obj) { +StatusWith<FailPoint::ModeOptions> FailPoint::parseBSON(const BSONObj& obj) { Mode mode = FailPoint::alwaysOn; ValType val = 0; const BSONElement modeElem(obj["mode"]); @@ -177,7 +176,6 @@ StatusWith<std::tuple<FailPoint::Mode, FailPoint::ValType, BSONObj>> FailPoint:: return {ErrorCodes::IllegalOperation, "When setting a failpoint, you must supply a 'mode'"}; } else if (modeElem.type() == String) { const std::string modeStr(modeElem.valuestr()); - if (modeStr == "off") { mode = FailPoint::off; } else if (modeStr == "alwaysOn") { @@ -255,7 +253,7 @@ StatusWith<std::tuple<FailPoint::Mode, FailPoint::ValType, BSONObj>> FailPoint:: data = obj["data"].Obj().getOwned(); } - return std::make_tuple(mode, val, data); + return ModeOptions{mode, val, data}; } BSONObj FailPoint::toBSON() const { diff --git a/src/mongo/util/fail_point.h b/src/mongo/util/fail_point.h index f82d85785a5..57ee76bca9d 100644 --- a/src/mongo/util/fail_point.h +++ b/src/mongo/util/fail_point.h @@ -44,22 +44,27 @@ namespace mongo { * deactivated, as well as embed temporary data into it. * * The fail point has a static instance, which is represented by a FailPoint - * object, and dynamic instance, which are all the threads in between - * shouldFailOpenBlock and shouldFailCloseBlock. + * object, and dynamic instance, represented by FailPoint::Scoped handles. * * Sample use: * // Declared somewhere: * FailPoint makeBadThingsHappen; * * // Somewhere in the code - * return false || MONGO_FAIL_POINT(makeBadThingsHappen); + * return false || MONGO_unlikely(makeBadThingsHappen.shouldFail()); * * or * * // Somewhere in the code - * MONGO_FAIL_POINT_BLOCK(makeBadThingsHappen, blockMakeBadThingsHappen) { - * const BSONObj& data = blockMakeBadThingsHappen.getData(); + * makeBadThingsHappen.execute([&](const BSONObj& data) { * // Do something + * }); + * + * // Another way to do it, where lambda isn't suitable, e.g. to cause an early return + * // of the enclosing function. + * if (auto sfp = makeBadThingsHappen.scoped(); MONGO_unlikely(sfp.isActive())) { + * const BSONObj& data = sfp.getData(); + * // Do something, including break, continue, return, etc... * } * * Invariants: @@ -69,13 +74,65 @@ namespace mongo { * 2. Client visible fail point states are read-only when active. */ class FailPoint { - FailPoint(const FailPoint&) = delete; - FailPoint& operator=(const FailPoint&) = delete; +private: + enum RetCode { fastOff = 0, slowOff, slowOn, userIgnored }; public: - typedef unsigned ValType; + using ValType = unsigned; enum Mode { off, alwaysOn, random, nTimes, skip }; - enum RetCode { fastOff = 0, slowOff, slowOn, userIgnored }; + + struct ModeOptions { + Mode mode; + ValType val; + BSONObj extra; + }; + + /** + * Helper class for making sure that FailPoint#shouldFailCloseBlock is called when + * FailPoint#shouldFailOpenBlock was called. + * + * Users don't create these. They are only used within the execute and executeIf + * functions and returned by the scoped() and scopedIf() functions. + */ + class Scoped { + public: + Scoped(FailPoint* failPoint, RetCode ret) + : _failPoint(failPoint), + _active(ret == FailPoint::slowOn), + _holdsRef(ret != FailPoint::fastOff) {} + + ~Scoped() { + if (_holdsRef) { + _failPoint->shouldFailCloseBlock(); + } + } + + Scoped(const Scoped&) = delete; + Scoped& operator=(const Scoped&) = delete; + + /** + * @return true if fail point is on. + * Calls to isActive should be placed inside MONGO_unlikely for performance. + */ + bool isActive() { + return _active; + } + + /** + * @return the data stored in the fail point. #isActive must be true + * before you can call this. + */ + const BSONObj& getData() const { + // Assert when attempting to get data without holding a ref. + fassert(16445, _holdsRef); + return _failPoint->getData(); + } + + private: + FailPoint* _failPoint; + bool _active; + bool _holdsRef; + }; /** * Explicitly resets the seed used for the PRNG in this thread. If not called on a thread, @@ -84,21 +141,26 @@ public: static void setThreadPRNGSeed(int32_t seed); /** - * Parses the FailPoint::Mode, FailPoint::ValType, and data BSONObj from the BSON. + * Parses the {Mode, ValType, BSONObj} from the BSON. */ - static StatusWith<std::tuple<Mode, ValType, BSONObj>> parseBSON(const BSONObj& obj); + static StatusWith<ModeOptions> parseBSON(const BSONObj& obj); FailPoint(); + FailPoint(const FailPoint&) = delete; + FailPoint& operator=(const FailPoint&) = delete; + /** * Note: This is not side-effect free - it can change the state to OFF after calling. - * Note: see MONGO_FAIL_POINT_BLOCK_IF for information on the passed callable + * Note: see `executeIf` for information on `pred`. + * + * Calls to shouldFail should be placed inside MONGO_unlikely for performance. * * @return true if fail point is active. */ - template <typename Callable = std::nullptr_t> - inline bool shouldFail(Callable&& cb = nullptr) { - RetCode ret = shouldFailOpenBlock(std::forward<Callable>(cb)); + template <typename Pred> + bool shouldFail(Pred&& pred) { + RetCode ret = shouldFailOpenBlock(std::forward<Pred>(pred)); if (MONGO_likely(ret == fastOff)) { return false; @@ -108,34 +170,11 @@ public: return ret == slowOn; } - /** - * Checks whether fail point is active and increments the reference counter without - * decrementing it. Must call shouldFailCloseBlock afterwards when the return value - * is not fastOff. Otherwise, this will remain read-only forever. - * - * Note: see MONGO_FAIL_POINT_BLOCK_IF for information on the passed callable - * - * @return slowOn if its active and needs to be closed - * userIgnored if its active and needs to be closed, but shouldn't be acted on - * slowOff if its disabled and needs to be closed - * fastOff if its disabled and doesn't need to be closed - */ - template <typename Callable = std::nullptr_t> - inline RetCode shouldFailOpenBlock(Callable&& cb = nullptr) { - if (MONGO_likely((_fpInfo.loadRelaxed() & ACTIVE_BIT) == 0)) { - return fastOff; - } - - return slowShouldFailOpenBlock(std::forward<Callable>(cb)); + bool shouldFail() { + return shouldFail(nullptr); } /** - * Decrements the reference counter. - * @see #shouldFailOpenBlock - */ - void shouldFailCloseBlock(); - - /** * Changes the settings of this fail point. This will turn off the fail point * and waits for all dynamic instances referencing this fail point to go away before * actually modifying the settings. @@ -157,143 +196,125 @@ public: * that can be referenced afterwards with #getData. Defaults to an empty * document. */ - void setMode(Mode mode, ValType val = 0, const BSONObj& extra = BSONObj()); + void setMode(Mode mode, ValType val = 0, BSONObj extra = {}); + void setMode(ModeOptions opt) { + setMode(std::move(opt.mode), std::move(opt.val), std::move(opt.extra)); + } /** * @returns a BSON object showing the current mode and data stored. */ BSONObj toBSON() const; -private: - static const ValType ACTIVE_BIT = 1 << 31; - static const ValType REF_COUNTER_MASK = ~ACTIVE_BIT; - - // Bit layout: - // 31: tells whether this fail point is active. - // 0~30: unsigned ref counter for active dynamic instances. - AtomicWord<unsigned> _fpInfo{0}; - - // Invariant: These should be read only if ACTIVE_BIT of _fpInfo is set. - Mode _mode{off}; - AtomicWord<int> _timesOrPeriod{0}; - BSONObj _data; - - // protects _mode, _timesOrPeriod, _data - mutable stdx::mutex _modMutex; - /** - * Enables this fail point. + * Create a Scoped from this FailPoint. + * Use the Scoped object to access failpoint data. */ - void enableFailPoint(); - + Scoped scoped() { + return scopedIf(nullptr); + } /** - * Disables this fail point. + * Create a Scoped from this FailPoint, only active when `pred(payload)` is true. + * See `executeIf`. Use the Scoped object to access failpoint data. */ - void disableFailPoint(); + template <typename Pred> + Scoped scopedIf(Pred&& pred) { + return Scoped(this, shouldFailOpenBlock(std::forward<Pred>(pred))); + } - /** - * slow path for #shouldFailOpenBlock - * - * If a callable is passed, and returns false, this will return userIgnored and avoid altering - * the mode in any way. The argument is the fail point payload. - */ - RetCode slowShouldFailOpenBlock(std::function<bool(const BSONObj&)> cb) noexcept; + template <typename F> + void execute(F&& f) { + return executeIf(f, nullptr); + } /** - * @return the stored BSONObj in this fail point. Note that this cannot be safely - * read if this fail point is off. + * The predicate `pred` should behave like a `bool pred(const BSONObj& payload)`. + * If `pred(payload)`, then `f(payload)` is executed. Otherwise, `f` is not + * executed and this FailPoint's mode is not altered (e.g. `nTimes` isn't consumed). */ - const BSONObj& getData() const; - - friend class ScopedFailPoint; -}; - -/** - * Helper class for making sure that FailPoint#shouldFailCloseBlock is called when - * FailPoint#shouldFailOpenBlock was called. This should only be used within the - * MONGO_FAIL_POINT_BLOCK macro. - */ -class ScopedFailPoint { - ScopedFailPoint(const ScopedFailPoint&) = delete; - ScopedFailPoint& operator=(const ScopedFailPoint&) = delete; + template <typename F, typename Pred> + void executeIf(F&& f, Pred&& pred) { + auto sfp = scopedIf(std::forward<Pred>(pred)); + if (MONGO_unlikely(sfp.isActive())) { + std::forward<F>(f)(sfp.getData()); + } + } -public: - template <typename Callable = std::nullptr_t> - ScopedFailPoint(FailPoint* failPoint, Callable&& cb = nullptr) : _failPoint(failPoint) { - FailPoint::RetCode ret = _failPoint->shouldFailOpenBlock(std::forward<Callable>(cb)); - _shouldClose = ret != FailPoint::fastOff; - _shouldRun = ret == FailPoint::slowOn; + void pauseWhileSet() { + while (MONGO_unlikely(shouldFail())) { + sleepmillis(100); + } } - ~ScopedFailPoint() { - if (_shouldClose) { - _failPoint->shouldFailCloseBlock(); + void pauseWhileSet(OperationContext* opCtx) { + while (MONGO_unlikely(shouldFail())) { + opCtx->sleepFor(Milliseconds(100)); } } +private: + void enable(); + void disable(); + /** - * @return true if fail point is on. This will be true at most once. + * Checks whether fail point is active and increments the reference counter without + * decrementing it. Must call shouldFailCloseBlock afterwards when the return value + * is not fastOff. Otherwise, this will remain read-only forever. + * + * Note: see `executeIf` for information on `pred`. + * + * @return slowOn if its active and needs to be closed + * userIgnored if its active and needs to be closed, but shouldn't be acted on + * slowOff if its disabled and needs to be closed + * fastOff if its disabled and doesn't need to be closed */ - inline bool isActive() { - if (!_shouldRun) { - return false; + template <typename Pred> + RetCode shouldFailOpenBlock(Pred&& pred) { + if (MONGO_likely((_fpInfo.loadRelaxed() & kActiveBit) == 0)) { + return fastOff; } - // We use this in a for loop to prevent iteration, thus flipping to inactive after the first - // time. - _shouldRun = false; - return true; + return slowShouldFailOpenBlock(std::forward<Pred>(pred)); + } + + RetCode shouldFailOpenBlock() { + return shouldFailOpenBlock(nullptr); } /** - * @return the data stored in the fail point. #isActive must be true - * before you can call this. + * Decrements the reference counter. + * @see #shouldFailOpenBlock */ - const BSONObj& getData() const { - // Assert when attempting to get data without incrementing ref counter. - fassert(16445, _shouldClose); - return _failPoint->getData(); - } + void shouldFailCloseBlock(); -private: - FailPoint* _failPoint; - bool _shouldRun; - bool _shouldClose; -}; + /** + * slow path for #shouldFailOpenBlock + * + * If a callable is passed, and returns false, this will return userIgnored and avoid altering + * the mode in any way. The argument is the fail point payload. + */ + RetCode slowShouldFailOpenBlock(std::function<bool(const BSONObj&)> cb) noexcept; -#define MONGO_FAIL_POINT(symbol) MONGO_unlikely(symbol.shouldFail()) + /** + * @return the stored BSONObj in this fail point. Note that this cannot be safely + * read if this fail point is off. + */ + const BSONObj& getData() const; -inline void MONGO_FAIL_POINT_PAUSE_WHILE_SET(FailPoint& failPoint) { - while (MONGO_FAIL_POINT(failPoint)) { - sleepmillis(100); - } -} + static const ValType kActiveBit = 1 << 31; -inline void MONGO_FAIL_POINT_PAUSE_WHILE_SET_OR_INTERRUPTED(OperationContext* opCtx, - FailPoint& failPoint) { - while (MONGO_FAIL_POINT(failPoint)) { - opCtx->sleepFor(Milliseconds(100)); - } -} + // Bit layout: + // 31: tells whether this fail point is active. + // 0~30: unsigned ref counter for active dynamic instances. + AtomicWord<std::uint32_t> _fpInfo{0}; -/** - * Macro for creating a fail point with block context. Also use this when - * you want to access the data stored in the fail point. - */ -#define MONGO_FAIL_POINT_BLOCK(symbol, blockSymbol) \ - for (mongo::ScopedFailPoint blockSymbol(&symbol); MONGO_unlikely(blockSymbol.isActive());) + // Invariant: These should be read only if kActiveBit of _fpInfo is set. + Mode _mode{off}; + AtomicWord<int> _timesOrPeriod{0}; + BSONObj _data; -/** - * Macro for creating a fail point with block context and a pre-flight condition. Also use this when - * you want to access the data stored in the fail point. - * - * Your passed in callable should take a const BSONObj& (the fail point payload) and return bool. - * If it returns true, you'll process the block as normal. If you return false, you'll exit the - * block without evaluating it and avoid altering the mode in any way (you won't consume nTimes for - * instance). - */ -#define MONGO_FAIL_POINT_BLOCK_IF(symbol, blockSymbol, ...) \ - for (mongo::ScopedFailPoint blockSymbol(&symbol, __VA_ARGS__); \ - MONGO_unlikely(blockSymbol.isActive());) + // protects _mode, _timesOrPeriod, _data + mutable stdx::mutex _modMutex; +}; } // namespace mongo diff --git a/src/mongo/util/fail_point_registry.cpp b/src/mongo/util/fail_point_registry.cpp index 158d6ab8023..f4dfb627a36 100644 --- a/src/mongo/util/fail_point_registry.cpp +++ b/src/mongo/util/fail_point_registry.cpp @@ -29,34 +29,32 @@ #include "mongo/util/fail_point_registry.h" +#include <fmt/format.h> + #include "mongo/bson/json.h" #include "mongo/util/fail_point_server_parameter_gen.h" #include "mongo/util/fail_point_service.h" -#include "mongo/util/map_util.h" -#include "mongo/util/str.h" namespace mongo { -constexpr auto kFailPointServerParameterPrefix = "failpoint."_sd; +using namespace fmt::literals; FailPointRegistry::FailPointRegistry() : _frozen(false) {} -Status FailPointRegistry::addFailPoint(const std::string& name, FailPoint* failPoint) { +Status FailPointRegistry::add(const std::string& name, FailPoint* failPoint) { if (_frozen) { return {ErrorCodes::CannotMutateObject, "Registry is already frozen"}; } - - if (_fpMap.count(name) > 0) { - return {ErrorCodes::Error(51006), - str::stream() << "Fail point already registered: " << name}; + auto [pos, ok] = _fpMap.insert({name, failPoint}); + if (!ok) { + return {ErrorCodes::Error(51006), "Fail point already registered: {}"_format(name)}; } - - _fpMap.insert(make_pair(name, failPoint)); return Status::OK(); } -FailPoint* FailPointRegistry::getFailPoint(const std::string& name) const { - return mapFindWithDefault(_fpMap, name, static_cast<FailPoint*>(nullptr)); +FailPoint* FailPointRegistry::find(const std::string& name) const { + auto iter = _fpMap.find(name); + return (iter == _fpMap.end()) ? nullptr : iter->second; } void FailPointRegistry::freeze() { @@ -64,17 +62,19 @@ void FailPointRegistry::freeze() { } void FailPointRegistry::registerAllFailPointsAsServerParameters() { - for (const auto& it : _fpMap) { + for (const auto& [name, ptr] : _fpMap) { // Intentionally leaked. - new FailPointServerParameter(it.first, ServerParameterType::kStartupOnly); + new FailPointServerParameter(name, ServerParameterType::kStartupOnly); } } +static constexpr auto kFailPointServerParameterPrefix = "failpoint."_sd; + FailPointServerParameter::FailPointServerParameter(StringData name, ServerParameterType spt) - : ServerParameter(kFailPointServerParameterPrefix.toString() + name.toString(), spt), - _data(getGlobalFailPointRegistry()->getFailPoint(name.toString())) { + : ServerParameter("{}{}"_format(kFailPointServerParameterPrefix, name), spt), + _data(globalFailPointRegistry().find(name.toString())) { invariant(name != "failpoint.*", "Failpoint prototype was auto-registered from IDL"); - invariant(_data != nullptr, str::stream() << "Unknown failpoint: " << name); + invariant(_data != nullptr, "Unknown failpoint: {}"_format(name)); } void FailPointServerParameter::append(OperationContext* opCtx, @@ -95,14 +95,7 @@ Status FailPointServerParameter::setFromString(const std::string& str) { if (!swParsedOptions.isOK()) { return swParsedOptions.getStatus(); } - - FailPoint::Mode mode; - FailPoint::ValType val; - BSONObj data; - std::tie(mode, val, data) = std::move(swParsedOptions.getValue()); - - _data->setMode(mode, val, data); - + _data->setMode(std::move(swParsedOptions.getValue())); return Status::OK(); } } // namespace mongo diff --git a/src/mongo/util/fail_point_registry.h b/src/mongo/util/fail_point_registry.h index 7446db95eb1..7a77b47faba 100644 --- a/src/mongo/util/fail_point_registry.h +++ b/src/mongo/util/fail_point_registry.h @@ -53,12 +53,12 @@ public: * 51006 - if the given name already exists in this registry. * CannotMutateObject - if this registry is already frozen. */ - Status addFailPoint(const std::string& name, FailPoint* failPoint); + Status add(const std::string& name, FailPoint* failPoint); /** - * @return the fail point object registered. Returns NULL if it was not registered. + * @return the fail point object registered, or nullptr if it was not registered. */ - FailPoint* getFailPoint(const std::string& name) const; + FailPoint* find(const std::string& name) const; /** * Freezes this registry from being modified. diff --git a/src/mongo/util/fail_point_service.cpp b/src/mongo/util/fail_point_service.cpp index 77068f82100..44d8c69ae65 100644 --- a/src/mongo/util/fail_point_service.cpp +++ b/src/mongo/util/fail_point_service.cpp @@ -38,49 +38,34 @@ namespace mongo { -using std::unique_ptr; - MONGO_FAIL_POINT_DEFINE(dummy); // used by tests in jstests/fail_point -unique_ptr<FailPointRegistry> _fpRegistry(nullptr); - -MONGO_INITIALIZER_GENERAL(FailPointRegistry, - MONGO_NO_PREREQUISITES, - ("BeginGeneralStartupOptionRegistration")) -(InitializerContext* context) { - _fpRegistry.reset(new FailPointRegistry()); - return Status::OK(); -} +MONGO_INITIALIZER_GROUP(FailPointRegistry, (), ("BeginStartupOptionHandling")); -MONGO_INITIALIZER_GENERAL(AllFailPointsRegistered, MONGO_NO_PREREQUISITES, MONGO_NO_DEPENDENTS) +MONGO_INITIALIZER_GENERAL(AllFailPointsRegistered, (), ()) (InitializerContext* context) { - _fpRegistry->freeze(); + globalFailPointRegistry().freeze(); return Status::OK(); } -FailPointRegistry* getGlobalFailPointRegistry() { - return _fpRegistry.get(); +FailPointRegistry& globalFailPointRegistry() { + static auto& p = *new FailPointRegistry(); + return p; } void setGlobalFailPoint(const std::string& failPointName, const BSONObj& cmdObj) { - FailPointRegistry* registry = getGlobalFailPointRegistry(); - FailPoint* failPoint = registry->getFailPoint(failPointName); + FailPoint* failPoint = globalFailPointRegistry().find(failPointName); if (failPoint == nullptr) uasserted(ErrorCodes::FailPointSetFailed, failPointName + " not found"); - FailPoint::Mode mode; - FailPoint::ValType val; - BSONObj data; - std::tie(mode, val, data) = uassertStatusOK(FailPoint::parseBSON(cmdObj)); - - failPoint->setMode(mode, val, data); + failPoint->setMode(uassertStatusOK(FailPoint::parseBSON(cmdObj))); warning() << "failpoint: " << failPointName << " set to: " << failPoint->toBSON(); } FailPointEnableBlock::FailPointEnableBlock(const std::string& failPointName) : _failPointName(failPointName) { - _failPoint = getGlobalFailPointRegistry()->getFailPoint(failPointName); + _failPoint = globalFailPointRegistry().find(failPointName); invariant(_failPoint != nullptr); _failPoint->setMode(FailPoint::alwaysOn); warning() << "failpoint: " << failPointName << " set to: " << _failPoint->toBSON(); @@ -88,7 +73,7 @@ FailPointEnableBlock::FailPointEnableBlock(const std::string& failPointName) FailPointEnableBlock::FailPointEnableBlock(const std::string& failPointName, const BSONObj& data) : _failPointName(failPointName) { - _failPoint = getGlobalFailPointRegistry()->getFailPoint(failPointName); + _failPoint = globalFailPointRegistry().find(failPointName); invariant(_failPoint != nullptr); _failPoint->setMode(FailPoint::alwaysOn, 0, data); warning() << "failpoint: " << failPointName << " set to: " << _failPoint->toBSON(); diff --git a/src/mongo/util/fail_point_service.h b/src/mongo/util/fail_point_service.h index 64c511942fc..8f7f3261b13 100644 --- a/src/mongo/util/fail_point_service.h +++ b/src/mongo/util/fail_point_service.h @@ -30,15 +30,13 @@ #pragma once #include "mongo/base/init.h" +#include "mongo/util/assert_util.h" #include "mongo/util/fail_point.h" #include "mongo/util/fail_point_registry.h" namespace mongo { -/** - * @return the global fail point registry. - */ -FailPointRegistry* getGlobalFailPointRegistry(); +FailPointRegistry& globalFailPointRegistry(); /** * Set a fail point in the global registry to a given value via BSON @@ -46,29 +44,27 @@ FailPointRegistry* getGlobalFailPointRegistry(); */ void setGlobalFailPoint(const std::string& failPointName, const BSONObj& cmdObj); +struct FailPointRegisterer { + FailPointRegisterer(const std::string& name, FailPoint* fp) { + uassertStatusOK(globalFailPointRegistry().add(name, fp)); + } +}; + /** * Convenience macro for defining a fail point. Must be used at namespace scope. * Note: that means never at local scope (inside functions) or class scope. * NOTE: Never use in header files, only sources. */ -#define MONGO_FAIL_POINT_DEFINE(fp) \ - ::mongo::FailPoint fp; \ - MONGO_INITIALIZER_GENERAL(fp, ("FailPointRegistry"), ("AllFailPointsRegistered")) \ - (::mongo::InitializerContext * context) { \ - return ::mongo::getGlobalFailPointRegistry()->addFailPoint(#fp, &fp); \ - } - -/** - * Convenience macro for declaring a fail point in a header. - */ -#define MONGO_FAIL_POINT_DECLARE(fp) extern ::mongo::FailPoint fp; +#define MONGO_FAIL_POINT_DEFINE(fp) \ + ::mongo::FailPoint fp; \ + ::mongo::FailPointRegisterer fp##failPointRegisterer(#fp, &fp); /** * Convenience class for enabling a failpoint and disabling it as this goes out of scope. */ class FailPointEnableBlock { public: - FailPointEnableBlock(const std::string& failPointName); + explicit FailPointEnableBlock(const std::string& failPointName); FailPointEnableBlock(const std::string& failPointName, const BSONObj& cmdObj); ~FailPointEnableBlock(); diff --git a/src/mongo/util/fail_point_test.cpp b/src/mongo/util/fail_point_test.cpp index 0a32ec45777..1880b0d18a1 100644 --- a/src/mongo/util/fail_point_test.cpp +++ b/src/mongo/util/fail_point_test.cpp @@ -46,7 +46,6 @@ using mongo::BSONObj; using mongo::FailPoint; using mongo::FailPointEnableBlock; -using mongo::getGlobalFailPointRegistry; namespace stdx = mongo::stdx; namespace mongo_test { @@ -60,7 +59,7 @@ TEST(FailPoint, AlwaysOn) { failPoint.setMode(FailPoint::alwaysOn); ASSERT(failPoint.shouldFail()); - MONGO_FAIL_POINT_BLOCK(failPoint, scopedFp) { + if (auto scopedFp = failPoint.scoped(); MONGO_unlikely(scopedFp.isActive())) { ASSERT(scopedFp.getData().isEmpty()); } @@ -85,11 +84,7 @@ TEST(FailPoint, NTimes) { TEST(FailPoint, BlockOff) { FailPoint failPoint; bool called = false; - - MONGO_FAIL_POINT_BLOCK(failPoint, scopedFp) { - called = true; - } - + failPoint.execute([&](const BSONObj&) { called = true; }); ASSERT_FALSE(called); } @@ -98,9 +93,7 @@ TEST(FailPoint, BlockAlwaysOn) { failPoint.setMode(FailPoint::alwaysOn); bool called = false; - MONGO_FAIL_POINT_BLOCK(failPoint, scopedFp) { - called = true; - } + failPoint.execute([&](const BSONObj&) { called = true; }); ASSERT(called); } @@ -111,9 +104,7 @@ TEST(FailPoint, BlockNTimes) { size_t counter = 0; for (size_t x = 0; x < 10; x++) { - MONGO_FAIL_POINT_BLOCK(failPoint, scopedFp) { - counter++; - } + failPoint.execute([&](auto&&...) { counter++; }); } ASSERT_EQUALS(1U, counter); @@ -125,9 +116,8 @@ TEST(FailPoint, BlockWithException) { bool threw = false; try { - MONGO_FAIL_POINT_BLOCK(failPoint, scopedFp) { - throw std::logic_error("BlockWithException threw"); - } + failPoint.execute( + [&](const BSONObj&) { throw std::logic_error("BlockWithException threw"); }); } catch (const std::logic_error&) { threw = true; } @@ -142,9 +132,7 @@ TEST(FailPoint, SetGetParam) { FailPoint failPoint; failPoint.setMode(FailPoint::alwaysOn, 0, BSON("x" << 20)); - MONGO_FAIL_POINT_BLOCK(failPoint, scopedFp) { - ASSERT_EQUALS(20, scopedFp.getData()["x"].numberInt()); - } + failPoint.execute([&](const BSONObj& data) { ASSERT_EQUALS(20, data["x"].numberInt()); }); } class FailPointStress : public mongo::unittest::Test { @@ -181,9 +169,7 @@ public: private: void blockTask() { while (true) { - MONGO_FAIL_POINT_BLOCK(_fp, scopedFp) { - const mongo::BSONObj& data = scopedFp.getData(); - + _fp.execute([](const BSONObj& data) { // Expanded ASSERT_EQUALS since the error is not being // printed out properly if (data["a"].numberInt() != 44) { @@ -191,7 +177,7 @@ private: << " - data: " << data << std::endl; ASSERT(false); } - } + }); stdx::lock_guard<stdx::mutex> lk(_mutex); if (_inShutdown) @@ -202,9 +188,7 @@ private: void blockWithExceptionTask() { while (true) { try { - MONGO_FAIL_POINT_BLOCK(_fp, scopedFp) { - const mongo::BSONObj& data = scopedFp.getData(); - + _fp.execute([](const BSONObj& data) { if (data["a"].numberInt() != 44) { mongo::error() << "blockWithExceptionTask thread detected anomaly" << " - data: " << data << std::endl; @@ -212,7 +196,7 @@ private: } throw std::logic_error("blockWithExceptionTask threw"); - } + }); } catch (const std::logic_error&) { } @@ -224,7 +208,7 @@ private: void simpleTask() { while (true) { - static_cast<void>(MONGO_FAIL_POINT(_fp)); + static_cast<void>(MONGO_unlikely(_fp.shouldFail())); stdx::lock_guard<stdx::mutex> lk(_mutex); if (_inShutdown) break; @@ -403,7 +387,7 @@ TEST(FailPoint, parseBSONValidDataSucceeds) { } TEST(FailPoint, FailPointBlockBasicTest) { - auto failPoint = getGlobalFailPointRegistry()->getFailPoint("dummy"); + auto failPoint = mongo::globalFailPointRegistry().find("dummy"); ASSERT_FALSE(failPoint->shouldFail()); @@ -418,33 +402,25 @@ TEST(FailPoint, FailPointBlockBasicTest) { TEST(FailPoint, FailPointBlockIfBasicTest) { FailPoint failPoint; failPoint.setMode(FailPoint::nTimes, 1, BSON("skip" << true)); - { bool hit = false; - - MONGO_FAIL_POINT_BLOCK_IF(failPoint, scopedFp, [&](const BSONObj& obj) { - hit = obj["skip"].trueValue(); - return false; - }) { - ASSERT(!"shouldn't get here"); - } - + failPoint.executeIf([](const BSONObj&) { ASSERT(!"shouldn't get here"); }, + [&hit](const BSONObj& obj) { + hit = obj["skip"].trueValue(); + return false; + }); ASSERT(hit); } - { bool hit = false; - - MONGO_FAIL_POINT_BLOCK_IF(failPoint, scopedFp, [](auto) { return true; }) { - hit = true; - ASSERT(!scopedFp.getData().isEmpty()); - } - + failPoint.executeIf( + [&hit](const BSONObj& data) { + hit = true; + ASSERT(!data.isEmpty()); + }, + [](const BSONObj&) { return true; }); ASSERT(hit); } - - MONGO_FAIL_POINT_BLOCK_IF(failPoint, scopedFp, [](auto) { return true; }) { - ASSERT(!"shouldn't get here"); - } + failPoint.executeIf([](auto&&) { ASSERT(!"shouldn't get here"); }, [](auto&&) { return true; }); } } // namespace mongo_test diff --git a/src/mongo/util/net/sock.cpp b/src/mongo/util/net/sock.cpp index b7cecfb3324..bd183130d5a 100644 --- a/src/mongo/util/net/sock.cpp +++ b/src/mongo/util/net/sock.cpp @@ -395,7 +395,7 @@ int Socket::_send(const char* data, int len, const char* context) { void Socket::send(const char* data, int len, const char* context) { while (len > 0) { int ret = -1; - if (MONGO_FAIL_POINT(throwSockExcep)) { + if (MONGO_unlikely(throwSockExcep.shouldFail())) { #if defined(_WIN32) WSASetLastError(WSAENETUNREACH); #else @@ -458,7 +458,7 @@ void Socket::send(const vector<pair<char*, int>>& data, const char* context) { while (meta.msg_iovlen > 0) { int ret = -1; - if (MONGO_FAIL_POINT(throwSockExcep)) { + if (MONGO_unlikely(throwSockExcep.shouldFail())) { #if defined(_WIN32) WSASetLastError(WSAENETUNREACH); #else @@ -491,7 +491,7 @@ void Socket::send(const vector<pair<char*, int>>& data, const char* context) { void Socket::recv(char* buf, int len) { while (len > 0) { int ret = -1; - if (MONGO_FAIL_POINT(throwSockExcep)) { + if (MONGO_unlikely(throwSockExcep.shouldFail())) { #if defined(_WIN32) WSASetLastError(WSAENETUNREACH); #else diff --git a/src/mongo/util/net/sock_test.cpp b/src/mongo/util/net/sock_test.cpp index d9b3ec8ee84..0d6d4c69017 100644 --- a/src/mongo/util/net/sock_test.cpp +++ b/src/mongo/util/net/sock_test.cpp @@ -208,7 +208,7 @@ const char kSocketFailPointName[] = "throwSockExcep"; class SocketFailPointTest : public unittest::Test { public: SocketFailPointTest() - : _failPoint(getGlobalFailPointRegistry()->getFailPoint(kSocketFailPointName)), + : _failPoint(globalFailPointRegistry().find(kSocketFailPointName)), _sockets(socketPair(SOCK_STREAM)) { ASSERT_TRUE(_failPoint != nullptr); ASSERT_TRUE(_sockets.first); diff --git a/src/mongo/util/net/ssl/detail/io.hpp b/src/mongo/util/net/ssl/detail/io.hpp index d6e376b00f0..f6ff6edd1d8 100644 --- a/src/mongo/util/net/ssl/detail/io.hpp +++ b/src/mongo/util/net/ssl/detail/io.hpp @@ -29,7 +29,7 @@ namespace ssl { namespace detail { // Failpoint to force small reads of data to exercise the SChannel buffering code -MONGO_FAIL_POINT_DECLARE(smallTLSReads); +extern ::mongo::FailPoint smallTLSReads; template <typename Stream, typename Operation> std::size_t io(Stream& next_layer, stream_core& core, const Operation& op, asio::error_code& ec) { @@ -42,7 +42,7 @@ std::size_t io(Stream& next_layer, stream_core& core, const Operation& op, asio: // the underlying transport. if (core.input_.size() == 0) { // Read tiny amounts of TLS data to test the SChannel buffering code - if (MONGO_FAIL_POINT(smallTLSReads)) { + if (MONGO_unlikely(smallTLSReads.shouldFail())) { core.input_ = asio::buffer(core.input_buffer_, next_layer.read_some( |