/** * Copyright (C) 2018-present MongoDB, Inc. * * This program is free software: you can redistribute it and/or modify * it under the terms of the Server Side Public License, version 1, * as published by MongoDB, Inc. * * This program is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the * Server Side Public License for more details. * * You should have received a copy of the Server Side Public License * along with this program. If not, see * . * * As a special exception, the copyright holders give permission to link the * code of portions of this program with the OpenSSL library under certain * conditions as described in each individual source file and distribute * linked combinations including the program with the OpenSSL library. You * must comply with the Server Side Public License in all respects for * all of the code used other than as permitted herein. If you modify file(s) * with this exception, you may extend this exception to your version of the * file(s), but you are not obligated to do so. If you do not wish to do so, * delete this exception statement from your version. If you delete this * exception statement from all source files in the program, then also delete * it in the license file. */ #define MONGO_LOGV2_DEFAULT_COMPONENT ::mongo::logv2::LogComponent::kCommand #include "mongo/platform/basic.h" #include "mongo/db/catalog/drop_collection.h" #include "mongo/db/audit.h" #include "mongo/db/catalog/index_catalog.h" #include "mongo/db/catalog/uncommitted_collections.h" #include "mongo/db/client.h" #include "mongo/db/concurrency/exception_util.h" #include "mongo/db/curop.h" #include "mongo/db/db_raii.h" #include "mongo/db/index_builds_coordinator.h" #include "mongo/db/repl/replication_coordinator.h" #include "mongo/db/s/collection_sharding_state.h" #include "mongo/db/server_options.h" #include "mongo/db/service_context.h" #include "mongo/db/views/view_catalog.h" #include "mongo/logv2/log.h" #include "mongo/util/fail_point.h" namespace mongo { MONGO_FAIL_POINT_DEFINE(hangDropCollectionBeforeLockAcquisition); MONGO_FAIL_POINT_DEFINE(hangDuringDropCollection); Status _checkNssAndReplState(OperationContext* opCtx, const CollectionPtr& coll) { if (!coll) { return Status(ErrorCodes::NamespaceNotFound, "ns not found"); } if (opCtx->writesAreReplicated() && !repl::ReplicationCoordinator::get(opCtx)->canAcceptWritesFor(opCtx, coll->ns())) { return Status(ErrorCodes::NotWritablePrimary, str::stream() << "Not primary while dropping collection " << coll->ns()); } return Status::OK(); } Status _dropView(OperationContext* opCtx, Database* db, const NamespaceString& collectionName, DropReply* reply) { if (!db) { Status status = Status(ErrorCodes::NamespaceNotFound, "ns not found"); audit::logDropView(opCtx->getClient(), collectionName, "", {}, status.code()); return status; } auto view = ViewCatalog::get(db)->lookupWithoutValidatingDurableViews(opCtx, collectionName.ns()); if (!view) { Status status = Status(ErrorCodes::NamespaceNotFound, "ns not found"); audit::logDropView(opCtx->getClient(), collectionName, "", {}, status.code()); return status; } // Validates the view or throws an "invalid view" error. ViewCatalog::get(db)->lookup(opCtx, collectionName.ns()); // Operations all lock system.views in the end to prevent deadlock. Lock::CollectionLock systemViewsLock(opCtx, db->getSystemViewsName(), MODE_X); if (MONGO_unlikely(hangDuringDropCollection.shouldFail())) { LOGV2(20330, "hangDuringDropCollection fail point enabled. Blocking until fail point is " "disabled."); hangDuringDropCollection.pauseWhileSet(); } AutoStatsTracker statsTracker( opCtx, collectionName, Top::LockType::NotLocked, AutoStatsTracker::LogMode::kUpdateCurOp, CollectionCatalog::get(opCtx)->getDatabaseProfileLevel(collectionName.db())); if (opCtx->writesAreReplicated() && !repl::ReplicationCoordinator::get(opCtx)->canAcceptWritesFor(opCtx, collectionName)) { return Status(ErrorCodes::NotWritablePrimary, str::stream() << "Not primary while dropping collection " << collectionName); } WriteUnitOfWork wunit(opCtx); audit::logDropView( opCtx->getClient(), collectionName, view->viewOn().ns(), view->pipeline(), ErrorCodes::OK); Status status = db->dropView(opCtx, collectionName); if (!status.isOK()) { return status; } wunit.commit(); reply->setNs(collectionName); return Status::OK(); } Status _abortIndexBuildsAndDrop(OperationContext* opCtx, AutoGetDb&& autoDb, const NamespaceString& startingNss, std::function&& dropFn, DropReply* reply, bool appendNs = true, boost::optional dropIfUUIDNotMatching = boost::none) { // We only need to hold an intent lock to send abort signals to the active index builder on this // collection. boost::optional optionalAutoDb(std::move(autoDb)); boost::optional collLock; collLock.emplace(opCtx, startingNss, MODE_IX); // Abandon the snapshot as the index catalog will compare the in-memory state to the disk state, // which may have changed when we released the collection lock temporarily. opCtx->recoveryUnit()->abandonSnapshot(); CollectionPtr coll = CollectionCatalog::get(opCtx)->lookupCollectionByNamespace(opCtx, startingNss); Status status = _checkNssAndReplState(opCtx, coll); if (!status.isOK()) { return status; } if (MONGO_unlikely(hangDuringDropCollection.shouldFail())) { LOGV2(518090, "hangDuringDropCollection fail point enabled. Blocking until fail point is " "disabled."); hangDuringDropCollection.pauseWhileSet(); } AutoStatsTracker statsTracker( opCtx, startingNss, Top::LockType::NotLocked, AutoStatsTracker::LogMode::kUpdateCurOp, CollectionCatalog::get(opCtx)->getDatabaseProfileLevel(startingNss.db())); IndexBuildsCoordinator* indexBuildsCoord = IndexBuildsCoordinator::get(opCtx); const UUID collectionUUID = coll->uuid(); if (dropIfUUIDNotMatching && collectionUUID == *dropIfUUIDNotMatching) { return Status::OK(); } const NamespaceStringOrUUID dbAndUUID{coll->ns().db().toString(), coll->uuid()}; const int numIndexes = coll->getIndexCatalog()->numIndexesTotal(opCtx); while (true) { // Save a copy of the namespace before yielding our locks. const NamespaceString collectionNs = coll->ns(); // Release locks before aborting index builds. The helper will acquire locks on our behalf. collLock = boost::none; optionalAutoDb = boost::none; // Send the abort signal to any active index builds on the collection. This waits until all // aborted index builds complete. indexBuildsCoord->abortCollectionIndexBuilds(opCtx, collectionNs, collectionUUID, str::stream() << "Collection " << collectionNs << "(" << collectionUUID << ") is being dropped"); // Take an exclusive lock to finish the collection drop. optionalAutoDb.emplace(opCtx, startingNss.db(), MODE_IX); collLock.emplace(opCtx, dbAndUUID, MODE_X); // Abandon the snapshot as the index catalog will compare the in-memory state to the // disk state, which may have changed when we released the collection lock temporarily. opCtx->recoveryUnit()->abandonSnapshot(); coll = CollectionCatalog::get(opCtx)->lookupCollectionByUUID(opCtx, collectionUUID); status = _checkNssAndReplState(opCtx, coll); if (!status.isOK()) { return status; } // Check if any new index builds were started while releasing the collection lock // temporarily, if so, we need to abort the new index builders. const bool abortAgain = indexBuildsCoord->inProgForCollection(collectionUUID); if (!abortAgain) { break; } } // It's possible for the given collection to be drop pending after obtaining the locks again, if // that is the case, then the collection is already registered to be dropped. Return early. const NamespaceString resolvedNss = coll->ns(); if (resolvedNss.isDropPendingNamespace()) { return Status::OK(); } // Serialize the drop with refreshes to prevent dropping a collection and creating the same // nss as a view while refreshing. CollectionShardingState::get(opCtx, resolvedNss)->checkShardVersionOrThrow(opCtx); invariant(coll->getIndexCatalog()->numIndexesInProgress(opCtx) == 0); status = dropFn(optionalAutoDb->getDb(), resolvedNss); if (!status.isOK()) { return status; } reply->setNIndexesWas(numIndexes); if (appendNs) { reply->setNs(resolvedNss); } return Status::OK(); } Status _dropCollectionForApplyOps(OperationContext* opCtx, Database* db, const NamespaceString& collectionName, const repl::OpTime& dropOpTime, DropCollectionSystemCollectionMode systemCollectionMode, DropReply* reply) { Lock::CollectionLock collLock(opCtx, collectionName, MODE_X); const CollectionPtr& coll = CollectionCatalog::get(opCtx)->lookupCollectionByNamespace(opCtx, collectionName); Status status = _checkNssAndReplState(opCtx, coll); if (!status.isOK()) { return status; } if (MONGO_unlikely(hangDuringDropCollection.shouldFail())) { LOGV2(20331, "hangDuringDropCollection fail point enabled. Blocking until fail point is " "disabled."); hangDuringDropCollection.pauseWhileSet(); } AutoStatsTracker statsTracker( opCtx, collectionName, Top::LockType::NotLocked, AutoStatsTracker::LogMode::kUpdateCurOp, CollectionCatalog::get(opCtx)->getDatabaseProfileLevel(collectionName.db())); WriteUnitOfWork wunit(opCtx); int numIndexes = coll->getIndexCatalog()->numIndexesTotal(opCtx); IndexBuildsCoordinator::get(opCtx)->assertNoIndexBuildInProgForCollection(coll->uuid()); status = systemCollectionMode == DropCollectionSystemCollectionMode::kDisallowSystemCollectionDrops ? db->dropCollection(opCtx, collectionName, dropOpTime) : db->dropCollectionEvenIfSystem(opCtx, collectionName, dropOpTime); if (!status.isOK()) { return status; } wunit.commit(); reply->setNIndexesWas(numIndexes); reply->setNs(collectionName); return Status::OK(); } Status _dropCollection(OperationContext* opCtx, const NamespaceString& nss, DropReply* reply, DropCollectionSystemCollectionMode systemCollectionMode, boost::optional dropIfUUIDNotMatching = boost::none) { // We rewrite drop of time-series buckets collection to drop of time-series view collection. // This ensures that such drop will delete both collections. const auto collectionName = nss.isTimeseriesBucketsCollection() ? nss.getTimeseriesViewNamespace() : nss; try { return writeConflictRetry(opCtx, "drop", collectionName.ns(), [&] { AutoGetDb autoDb(opCtx, collectionName.db(), MODE_IX); auto db = autoDb.getDb(); if (!db) { return Status(ErrorCodes::NamespaceNotFound, "ns not found"); } if (CollectionCatalog::get(opCtx)->lookupCollectionByNamespace(opCtx, collectionName)) { return _abortIndexBuildsAndDrop( opCtx, std::move(autoDb), collectionName, [opCtx, systemCollectionMode](Database* db, const NamespaceString& resolvedNs) { WriteUnitOfWork wuow(opCtx); auto status = systemCollectionMode == DropCollectionSystemCollectionMode::kDisallowSystemCollectionDrops ? db->dropCollection(opCtx, resolvedNs) : db->dropCollectionEvenIfSystem(opCtx, resolvedNs); if (!status.isOK()) { return status; } wuow.commit(); return Status::OK(); }, reply, true /* appendNs */, dropIfUUIDNotMatching); } auto dropTimeseries = [opCtx, &autoDb, &collectionName, &reply]( const NamespaceString& bucketNs, bool dropView) { return _abortIndexBuildsAndDrop( opCtx, std::move(autoDb), bucketNs, [opCtx, dropView, &collectionName, &reply](Database* db, const NamespaceString& bucketsNs) { if (dropView) { // Take a MODE_X lock when dropping timeseries view. This is to prevent // a concurrent create collection on the same namespace that will // reserve an OpTime before this drop. We already hold a MODE_X lock on // the bucket collection inside '_abortIndexBuildsAndDrop' above. When // taking both these locks it needs to happen in this order to prevent a // deadlock. Lock::CollectionLock viewLock(opCtx, collectionName, MODE_X); auto status = _dropView(opCtx, db, collectionName, reply); if (!status.isOK()) { return status; } } // Drop the buckets collection in its own writeConflictRetry so that if // it throws a WCE, only the buckets collection drop is retried. writeConflictRetry(opCtx, "drop", bucketsNs.ns(), [opCtx, db, &bucketsNs] { WriteUnitOfWork wuow(opCtx); db->dropCollectionEvenIfSystem(opCtx, bucketsNs).ignore(); wuow.commit(); }); return Status::OK(); }, reply, false /* appendNs */); }; auto view = ViewCatalog::get(db)->lookupWithoutValidatingDurableViews( opCtx, collectionName.ns()); if (!view) { // Timeseries bucket collection may exist even without the view. If that is the case // delete it. auto bucketsNs = collectionName.makeTimeseriesBucketsNamespace(); if (CollectionCatalog::get(opCtx)->lookupCollectionByNamespace(opCtx, bucketsNs)) { return dropTimeseries(bucketsNs, false); } Status status = Status(ErrorCodes::NamespaceNotFound, "ns not found"); audit::logDropView(opCtx->getClient(), collectionName, "", {}, status.code()); return status; } if (view->timeseries() && CollectionCatalog::get(opCtx)->lookupCollectionByNamespace(opCtx, view->viewOn())) { return dropTimeseries(view->viewOn(), true); } // Take a MODE_X lock when dropping a view. This is to prevent a concurrent create // collection on the same namespace that will reserve an OpTime before this drop. Lock::CollectionLock viewLock(opCtx, collectionName, MODE_X); return _dropView(opCtx, db, collectionName, reply); }); } catch (ExceptionFor&) { // The shell requires that NamespaceNotFound error codes return the "ns not found" // string. return Status(ErrorCodes::NamespaceNotFound, "ns not found"); } } Status dropCollection(OperationContext* opCtx, const NamespaceString& nss, DropReply* reply, DropCollectionSystemCollectionMode systemCollectionMode) { if (!serverGlobalParams.quiet.load()) { LOGV2(518070, "CMD: drop", logAttrs(nss)); } if (MONGO_unlikely(hangDropCollectionBeforeLockAcquisition.shouldFail())) { LOGV2(518080, "Hanging drop collection before lock acquisition while fail point is set"); hangDropCollectionBeforeLockAcquisition.pauseWhileSet(); } return _dropCollection(opCtx, nss, reply, systemCollectionMode); } Status dropCollectionIfUUIDNotMatching(OperationContext* opCtx, const NamespaceString& ns, const UUID& expectedUUID) { AutoGetDb autoDb(opCtx, ns.db(), MODE_IX); if (autoDb.getDb()) { { Lock::CollectionLock collLock(opCtx, ns, MODE_IS); const auto coll = CollectionCatalog::get(opCtx)->lookupCollectionByNamespace(opCtx, ns); if (!coll || coll->uuid() == expectedUUID) { return Status::OK(); } } DropReply repl; return _dropCollection(opCtx, ns, &repl, DropCollectionSystemCollectionMode::kDisallowSystemCollectionDrops, expectedUUID); } return Status::OK(); } Status dropCollectionForApplyOps(OperationContext* opCtx, const NamespaceString& collectionName, const repl::OpTime& dropOpTime, DropCollectionSystemCollectionMode systemCollectionMode) { if (!serverGlobalParams.quiet.load()) { LOGV2(20332, "CMD: drop", logAttrs(collectionName)); } if (MONGO_unlikely(hangDropCollectionBeforeLockAcquisition.shouldFail())) { LOGV2(20333, "Hanging drop collection before lock acquisition while fail point is set"); hangDropCollectionBeforeLockAcquisition.pauseWhileSet(); } return writeConflictRetry(opCtx, "drop", collectionName.ns(), [&] { AutoGetDb autoDb(opCtx, collectionName.db(), MODE_IX); Database* db = autoDb.getDb(); if (!db) { return Status(ErrorCodes::NamespaceNotFound, "ns not found"); } const CollectionPtr& coll = CollectionCatalog::get(opCtx)->lookupCollectionByNamespace(opCtx, collectionName); DropReply unusedReply; if (!coll) { Lock::CollectionLock viewLock(opCtx, collectionName, MODE_IX); return _dropView(opCtx, db, collectionName, &unusedReply); } else { return _dropCollectionForApplyOps( opCtx, db, collectionName, dropOpTime, systemCollectionMode, &unusedReply); } }); } } // namespace mongo