/** * Copyright (C) 2016 MongoDB Inc. * * This program is free software: you can redistribute it and/or modify * it under the terms of the GNU Affero General Public License, version 3, * as published by the Free Software Foundation. * * 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 * GNU Affero General Public License for more details. * * You should have received a copy of the GNU Affero General 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 GNU Affero General 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_LOG_DEFAULT_COMPONENT ::mongo::logger::LogComponent::kStorage #include "mongo/platform/basic.h" #include #include #include #include #include #include #include #include "mongo/base/checked_cast.h" #include "mongo/base/init.h" #include "mongo/base/initializer.h" #include "mongo/base/status.h" #include "mongo/client/replica_set_monitor.h" #include "mongo/config.h" #include "mongo/db/audit.h" #include "mongo/db/auth/auth_index_d.h" #include "mongo/db/auth/authorization_manager.h" #include "mongo/db/auth/authorization_manager_global.h" #include "mongo/db/catalog/collection.h" #include "mongo/db/catalog/database.h" #include "mongo/db/catalog/database_catalog_entry.h" #include "mongo/db/catalog/database_holder.h" #include "mongo/db/catalog/index_catalog.h" #include "mongo/db/catalog/index_key_validate.h" #include "mongo/db/client.h" #include "mongo/db/clientcursor.h" #include "mongo/db/commands/feature_compatibility_version.h" #include "mongo/db/concurrency/d_concurrency.h" #include "mongo/db/concurrency/lock_state.h" #include "mongo/db/db_raii.h" #include "mongo/db/dbdirectclient.h" #include "mongo/db/dbhelpers.h" #include "mongo/db/dbmessage.h" #include "mongo/db/diag_log.h" #include "mongo/db/exec/working_set_common.h" #include "mongo/db/ftdc/ftdc_mongod.h" #include "mongo/db/index_names.h" #include "mongo/db/index_rebuilder.h" #include "mongo/db/initialize_server_global_state.h" #include "mongo/db/initialize_snmp.h" #include "mongo/db/introspect.h" #include "mongo/db/json.h" #include "mongo/db/log_process_details.h" #include "mongo/db/logical_clock.h" #include "mongo/db/logical_time_metadata_hook.h" #include "mongo/db/mongod_options.h" #include "mongo/db/op_observer_impl.h" #include "mongo/db/operation_context.h" #include "mongo/db/query/internal_plans.h" #include "mongo/db/repair_database.h" #include "mongo/db/repl/oplog.h" #include "mongo/db/repl/repl_settings.h" #include "mongo/db/repl/replication_coordinator_external_state_impl.h" #include "mongo/db/repl/replication_coordinator_global.h" #include "mongo/db/repl/replication_coordinator_impl.h" #include "mongo/db/repl/replication_process.h" #include "mongo/db/repl/storage_interface_impl.h" #include "mongo/db/repl/topology_coordinator_impl.h" #include "mongo/db/s/balancer/balancer.h" #include "mongo/db/s/sharding_initialization_mongod.h" #include "mongo/db/s/sharding_state.h" #include "mongo/db/s/sharding_state_recovery.h" #include "mongo/db/s/type_shard_identity.h" #include "mongo/db/server_options.h" #include "mongo/db/server_parameters.h" #include "mongo/db/service_context.h" #include "mongo/db/service_context_d.h" #include "mongo/db/service_entry_point_mongod.h" #include "mongo/db/startup_warnings_mongod.h" #include "mongo/db/stats/counters.h" #include "mongo/db/storage/encryption_hooks.h" #include "mongo/db/storage/mmap_v1/mmap_v1_options.h" #include "mongo/db/storage/storage_engine.h" #include "mongo/db/storage/storage_options.h" #include "mongo/db/ttl.h" #include "mongo/db/wire_version.h" #include "mongo/executor/network_connection_hook.h" #include "mongo/executor/network_interface_factory.h" #include "mongo/platform/process_id.h" #include "mongo/rpc/metadata/egress_metadata_hook_list.h" #include "mongo/s/client/shard_registry.h" #include "mongo/s/grid.h" #include "mongo/s/sharding_initialization.h" #include "mongo/scripting/dbdirectclient_factory.h" #include "mongo/scripting/engine.h" #include "mongo/stdx/future.h" #include "mongo/stdx/memory.h" #include "mongo/stdx/thread.h" #include "mongo/transport/transport_layer_legacy.h" #include "mongo/util/assert_util.h" #include "mongo/util/background.h" #include "mongo/util/cmdline_utils/censor_cmdline.h" #include "mongo/util/concurrency/idle_thread_block.h" #include "mongo/util/concurrency/thread_name.h" #include "mongo/util/exception_filter_win32.h" #include "mongo/util/exit.h" #include "mongo/util/fail_point_service.h" #include "mongo/util/fast_clock_source_factory.h" #include "mongo/util/log.h" #include "mongo/util/net/listen.h" #include "mongo/util/net/ssl_manager.h" #include "mongo/util/ntservice.h" #include "mongo/util/options_parser/startup_options.h" #include "mongo/util/quick_exit.h" #include "mongo/util/ramlog.h" #include "mongo/util/scopeguard.h" #include "mongo/util/signal_handlers.h" #include "mongo/util/stacktrace.h" #include "mongo/util/startup_test.h" #include "mongo/util/text.h" #include "mongo/util/time_support.h" #include "mongo/util/version.h" #if !defined(_WIN32) #include #endif namespace mongo { using std::unique_ptr; using std::cout; using std::cerr; using std::endl; using std::list; using std::string; using std::stringstream; using std::vector; using logger::LogComponent; extern int diagLogging; namespace { const NamespaceString startupLogCollectionName("local.startup_log"); const NamespaceString kSystemReplSetCollection("local.system.replset"); #ifdef _WIN32 ntservice::NtServiceDefaultStrings defaultServiceStrings = { L"MongoDB", L"MongoDB", L"MongoDB Server"}; #endif Timer startupSrandTimer; void logStartup(OperationContext* opCtx) { BSONObjBuilder toLog; stringstream id; id << getHostNameCached() << "-" << jsTime().asInt64(); toLog.append("_id", id.str()); toLog.append("hostname", getHostNameCached()); toLog.appendTimeT("startTime", time(0)); toLog.append("startTimeLocal", dateToCtimeString(Date_t::now())); toLog.append("cmdLine", serverGlobalParams.parsedOpts); toLog.append("pid", ProcessId::getCurrent().asLongLong()); BSONObjBuilder buildinfo(toLog.subobjStart("buildinfo")); VersionInfoInterface::instance().appendBuildInfo(&buildinfo); appendStorageEngineList(&buildinfo); buildinfo.doneFast(); BSONObj o = toLog.obj(); Lock::GlobalWrite lk(opCtx); AutoGetOrCreateDb autoDb(opCtx, startupLogCollectionName.db(), mongo::MODE_X); Database* db = autoDb.getDb(); Collection* collection = db->getCollection(opCtx, startupLogCollectionName); WriteUnitOfWork wunit(opCtx); if (!collection) { BSONObj options = BSON("capped" << true << "size" << 10 * 1024 * 1024); repl::UnreplicatedWritesBlock uwb(opCtx); uassertStatusOK(userCreateNS(opCtx, db, startupLogCollectionName.ns(), options)); collection = db->getCollection(opCtx, startupLogCollectionName); } invariant(collection); OpDebug* const nullOpDebug = nullptr; uassertStatusOK(collection->insertDocument(opCtx, o, nullOpDebug, false)); wunit.commit(); } void checkForIdIndexes(OperationContext* opCtx, Database* db) { if (db->name() == "local") { // we do not need an _id index on anything in the local database return; } list collections; db->getDatabaseCatalogEntry()->getCollectionNamespaces(&collections); // for each collection, ensure there is a $_id_ index for (list::iterator i = collections.begin(); i != collections.end(); ++i) { const string& collectionName = *i; NamespaceString ns(collectionName); if (ns.isSystem()) continue; Collection* coll = db->getCollection(opCtx, collectionName); if (!coll) continue; if (coll->getIndexCatalog()->findIdIndex(opCtx)) continue; log() << "WARNING: the collection '" << *i << "' lacks a unique index on _id." << " This index is needed for replication to function properly" << startupWarningsLog; log() << "\t To fix this, you need to create a unique index on _id." << " See http://dochub.mongodb.org/core/build-replica-set-indexes" << startupWarningsLog; } } /** * Checks if this server was started without --replset but has a config in local.system.replset * (meaning that this is probably a replica set member started in stand-alone mode). * * @returns the number of documents in local.system.replset or 0 if this was started with * --replset. */ unsigned long long checkIfReplMissingFromCommandLine(OperationContext* opCtx) { // This is helpful for the query below to work as you can't open files when readlocked Lock::GlobalWrite lk(opCtx); if (!repl::getGlobalReplicationCoordinator()->getSettings().usingReplSets()) { DBDirectClient c(opCtx); return c.count(kSystemReplSetCollection.ns()); } return 0; } /** * Check that the oplog is capped, and abort the process if it is not. * Caller must lock DB before calling this function. */ void checkForCappedOplog(OperationContext* opCtx, Database* db) { const NamespaceString oplogNss(repl::rsOplogName); invariant(opCtx->lockState()->isDbLockedForMode(oplogNss.db(), MODE_IS)); Collection* oplogCollection = db->getCollection(opCtx, oplogNss); if (oplogCollection && !oplogCollection->isCapped()) { severe() << "The oplog collection " << oplogNss << " is not capped; a capped oplog is a requirement for replication to function."; fassertFailedNoTrace(40115); } } void repairDatabasesAndCheckVersion(OperationContext* opCtx) { LOG(1) << "enter repairDatabases (to check pdfile version #)"; Lock::GlobalWrite lk(opCtx); vector dbNames; StorageEngine* storageEngine = opCtx->getServiceContext()->getGlobalStorageEngine(); storageEngine->listDatabases(&dbNames); // Repair all databases first, so that we do not try to open them if they are in bad shape if (storageGlobalParams.repair) { invariant(!storageGlobalParams.readOnly); for (vector::const_iterator i = dbNames.begin(); i != dbNames.end(); ++i) { const string dbName = *i; LOG(1) << " Repairing database: " << dbName; fassert(18506, repairDatabase(opCtx, storageEngine, dbName)); } } const repl::ReplSettings& replSettings = repl::getGlobalReplicationCoordinator()->getSettings(); if (!storageGlobalParams.readOnly) { // We open the "local" database before calling checkIfReplMissingFromCommandLine() to ensure // the in-memory catalog entries for the 'kSystemReplSetCollection' collection have been // populated if the collection exists. If the "local" database didn't exist at this point // yet, then it will be created. If the mongod is running in a read-only mode, then it is // fine to not open the "local" database and populate the catalog entries because we won't // attempt to drop the temporary collections anyway. Lock::DBLock dbLock(opCtx, kSystemReplSetCollection.db(), MODE_X); dbHolder().openDb(opCtx, kSystemReplSetCollection.db()); } // On replica set members we only clear temp collections on DBs other than "local" during // promotion to primary. On pure slaves, they are only cleared when the oplog tells them // to. The local DB is special because it is not replicated. See SERVER-10927 for more // details. const bool shouldClearNonLocalTmpCollections = !(checkIfReplMissingFromCommandLine(opCtx) || replSettings.usingReplSets() || replSettings.isSlave()); for (vector::const_iterator i = dbNames.begin(); i != dbNames.end(); ++i) { const string dbName = *i; LOG(1) << " Recovering database: " << dbName; Database* db = dbHolder().openDb(opCtx, dbName); invariant(db); // First thing after opening the database is to check for file compatibility, // otherwise we might crash if this is a deprecated format. auto status = db->getDatabaseCatalogEntry()->currentFilesCompatible(opCtx); if (!status.isOK()) { if (status.code() == ErrorCodes::CanRepairToDowngrade) { // Convert CanRepairToDowngrade statuses to MustUpgrade statuses to avoid logging a // potentially confusing and inaccurate message. // // TODO SERVER-24097: Log a message informing the user that they can start the // current version of mongod with --repair and then proceed with normal startup. status = {ErrorCodes::MustUpgrade, status.reason()}; } severe() << "Unable to start mongod due to an incompatibility with the data files and" " this version of mongod: " << redact(status); severe() << "Please consult our documentation when trying to downgrade to a previous" " major release"; quickExit(EXIT_NEED_UPGRADE); return; } // Check if admin.system.version contains an invalid featureCompatibilityVersion. // If a valid featureCompatibilityVersion is present, cache it as a server parameter. if (dbName == "admin") { if (Collection* versionColl = db->getCollection(opCtx, FeatureCompatibilityVersion::kCollection)) { BSONObj featureCompatibilityVersion; if (Helpers::findOne(opCtx, versionColl, BSON("_id" << FeatureCompatibilityVersion::kParameterName), featureCompatibilityVersion)) { auto version = FeatureCompatibilityVersion::parse(featureCompatibilityVersion); if (!version.isOK()) { severe() << version.getStatus(); fassertFailedNoTrace(40283); } serverGlobalParams.featureCompatibility.version.store(version.getValue()); } } } // Major versions match, check indexes const NamespaceString systemIndexes(db->name(), "system.indexes"); Collection* coll = db->getCollection(opCtx, systemIndexes); auto exec = InternalPlanner::collectionScan( opCtx, systemIndexes.ns(), coll, PlanExecutor::NO_YIELD); BSONObj index; PlanExecutor::ExecState state; while (PlanExecutor::ADVANCED == (state = exec->getNext(&index, NULL))) { const BSONObj key = index.getObjectField("key"); const string plugin = IndexNames::findPluginName(key); if (db->getDatabaseCatalogEntry()->isOlderThan24(opCtx)) { if (IndexNames::existedBefore24(plugin)) { continue; } log() << "Index " << index << " claims to be of type '" << plugin << "', " << "which is either invalid or did not exist before v2.4. " << "See the upgrade section: " << "http://dochub.mongodb.org/core/upgrade-2.4" << startupWarningsLog; } if (index["v"].isNumber() && index["v"].numberInt() == 0) { log() << "WARNING: The index: " << index << " was created with the deprecated" << " v:0 format. This format will not be supported in a future release." << startupWarningsLog; log() << "\t To fix this, you need to rebuild this index." << " For instructions, see http://dochub.mongodb.org/core/rebuild-v0-indexes" << startupWarningsLog; } } // Non-yielding collection scans from InternalPlanner will never error. invariant(PlanExecutor::IS_EOF == state); if (replSettings.usingReplSets()) { // We only care about the _id index if we are in a replset checkForIdIndexes(opCtx, db); // Ensure oplog is capped (mmap does not guarantee order of inserts on noncapped // collections) if (db->name() == "local") { checkForCappedOplog(opCtx, db); } } if (!storageGlobalParams.readOnly && (shouldClearNonLocalTmpCollections || dbName == "local")) { db->clearTmpCollections(opCtx); } } LOG(1) << "done repairDatabases"; } void _initWireSpec() { WireSpec& spec = WireSpec::instance(); spec.isInternalClient = true; } MONGO_FP_DECLARE(shutdownAtStartup); ExitCode _initAndListen(int listenPort) { Client::initThread("initandlisten"); _initWireSpec(); auto globalServiceContext = checked_cast(getGlobalServiceContext()); globalServiceContext->setFastClockSource(FastClockSourceFactory::create(Milliseconds(10))); globalServiceContext->setOpObserver(stdx::make_unique()); DBDirectClientFactory::get(globalServiceContext) .registerImplementation([](OperationContext* opCtx) { return std::unique_ptr(new DBDirectClient(opCtx)); }); const repl::ReplSettings& replSettings = repl::getGlobalReplicationCoordinator()->getSettings(); { ProcessId pid = ProcessId::getCurrent(); LogstreamBuilder l = log(LogComponent::kControl); l << "MongoDB starting : pid=" << pid << " port=" << serverGlobalParams.port << " dbpath=" << storageGlobalParams.dbpath; if (replSettings.isMaster()) l << " master=" << replSettings.isMaster(); if (replSettings.isSlave()) l << " slave=" << (int)replSettings.isSlave(); const bool is32bit = sizeof(int*) == 4; l << (is32bit ? " 32" : " 64") << "-bit host=" << getHostNameCached() << endl; } DEV log(LogComponent::kControl) << "DEBUG build (which is slower)" << endl; #if defined(_WIN32) VersionInfoInterface::instance().logTargetMinOS(); #endif logProcessDetails(); globalServiceContext->createLockFile(); transport::TransportLayerLegacy::Options options; options.port = listenPort; options.ipList = serverGlobalParams.bind_ip; globalServiceContext->setServiceEntryPoint( stdx::make_unique(globalServiceContext->getTransportLayer())); // Create, start, and attach the TL auto transportLayer = stdx::make_unique( options, globalServiceContext->getServiceEntryPoint()); auto res = transportLayer->setup(); if (!res.isOK()) { error() << "Failed to set up listener: " << res; return EXIT_NET_ERROR; } globalServiceContext->initializeGlobalStorageEngine(); #ifdef MONGO_CONFIG_WIREDTIGER_ENABLED if (EncryptionHooks::get(getGlobalServiceContext())->restartRequired()) { exitCleanly(EXIT_CLEAN); } #endif // Warn if we detect configurations for multiple registered storage engines in // the same configuration file/environment. if (serverGlobalParams.parsedOpts.hasField("storage")) { BSONElement storageElement = serverGlobalParams.parsedOpts.getField("storage"); invariant(storageElement.isABSONObj()); BSONObj storageParamsObj = storageElement.Obj(); BSONObjIterator i = storageParamsObj.begin(); while (i.more()) { BSONElement e = i.next(); // Ignore if field name under "storage" matches current storage engine. if (storageGlobalParams.engine == e.fieldName()) { continue; } // Warn if field name matches non-active registered storage engine. if (globalServiceContext->isRegisteredStorageEngine(e.fieldName())) { warning() << "Detected configuration for non-active storage engine " << e.fieldName() << " when current storage engine is " << storageGlobalParams.engine; } } } if (!globalServiceContext->getGlobalStorageEngine()->getSnapshotManager()) { if (moe::startupOptionsParsed.count("replication.enableMajorityReadConcern") && moe::startupOptionsParsed["replication.enableMajorityReadConcern"].as()) { // Note: we are intentionally only erroring if the user explicitly requested that we // enable majority read concern. We do not error if the they are implicitly enabled for // CSRS because a required step in the upgrade procedure can involve an mmapv1 node in // the CSRS in the REMOVED state. This is handled by the TopologyCoordinator. invariant(replSettings.isMajorityReadConcernEnabled()); severe() << "Majority read concern requires a storage engine that supports" << " snapshots, such as wiredTiger. " << storageGlobalParams.engine << " does not support snapshots."; exitCleanly(EXIT_BADOPTIONS); } } logMongodStartupWarnings(storageGlobalParams, serverGlobalParams); { stringstream ss; ss << endl; ss << "*********************************************************************" << endl; ss << " ERROR: dbpath (" << storageGlobalParams.dbpath << ") does not exist." << endl; ss << " Create this directory or give existing directory in --dbpath." << endl; ss << " See http://dochub.mongodb.org/core/startingandstoppingmongo" << endl; ss << "*********************************************************************" << endl; uassert(10296, ss.str().c_str(), boost::filesystem::exists(storageGlobalParams.dbpath)); } { stringstream ss; ss << "repairpath (" << storageGlobalParams.repairpath << ") does not exist"; uassert(12590, ss.str().c_str(), boost::filesystem::exists(storageGlobalParams.repairpath)); } initializeSNMP(); if (!storageGlobalParams.readOnly) { boost::filesystem::remove_all(storageGlobalParams.dbpath + "/_tmp/"); } if (mmapv1GlobalOptions.journalOptions & MMAPV1Options::JournalRecoverOnly) return EXIT_NET_ERROR; if (mongodGlobalParams.scriptingEnabled) { ScriptEngine::setup(); } auto startupOpCtx = globalServiceContext->makeOperationContext(&cc()); repairDatabasesAndCheckVersion(startupOpCtx.get()); if (storageGlobalParams.upgrade) { log() << "finished checking dbs"; exitCleanly(EXIT_CLEAN); } uassertStatusOK(getGlobalAuthorizationManager()->initialize(startupOpCtx.get())); /* this is for security on certain platforms (nonce generation) */ srand((unsigned)(curTimeMicros64() ^ startupSrandTimer.micros())); AuthorizationManager* globalAuthzManager = getGlobalAuthorizationManager(); if (globalAuthzManager->shouldValidateAuthSchemaOnStartup()) { Status status = authindex::verifySystemIndexes(startupOpCtx.get()); if (!status.isOK()) { log() << redact(status); exitCleanly(EXIT_NEED_UPGRADE); } // SERVER-14090: Verify that auth schema version is schemaVersion26Final. int foundSchemaVersion; status = globalAuthzManager->getAuthorizationVersion(startupOpCtx.get(), &foundSchemaVersion); if (!status.isOK()) { log() << "Auth schema version is incompatible: " << "User and role management commands require auth data to have " << "at least schema version " << AuthorizationManager::schemaVersion26Final << " but startup could not verify schema version: " << status; exitCleanly(EXIT_NEED_UPGRADE); } if (foundSchemaVersion < AuthorizationManager::schemaVersion26Final) { log() << "Auth schema version is incompatible: " << "User and role management commands require auth data to have " << "at least schema version " << AuthorizationManager::schemaVersion26Final << " but found " << foundSchemaVersion << ". In order to upgrade " << "the auth schema, first downgrade MongoDB binaries to version " << "2.6 and then run the authSchemaUpgrade command."; exitCleanly(EXIT_NEED_UPGRADE); } } else if (globalAuthzManager->isAuthEnabled()) { error() << "Auth must be disabled when starting without auth schema validation"; exitCleanly(EXIT_BADOPTIONS); } else { // If authSchemaValidation is disabled and server is running without auth, // warn the user and continue startup without authSchema metadata checks. log() << startupWarningsLog; log() << "** WARNING: Startup auth schema validation checks are disabled for the " "database." << startupWarningsLog; log() << "** This mode should only be used to manually repair corrupted auth " "data." << startupWarningsLog; } auto shardingInitialized = uassertStatusOK(ShardingState::get(startupOpCtx.get()) ->initializeShardingAwarenessIfNeeded(startupOpCtx.get())); if (shardingInitialized) { reloadShardRegistryUntilSuccess(startupOpCtx.get()); } if (!storageGlobalParams.readOnly) { logStartup(startupOpCtx.get()); startFTDC(); restartInProgressIndexesFromLastShutdown(startupOpCtx.get()); if (serverGlobalParams.clusterRole == ClusterRole::ShardServer) { // Note: For replica sets, ShardingStateRecovery happens on transition to primary. if (!repl::getGlobalReplicationCoordinator()->isReplEnabled()) { uassertStatusOK(ShardingStateRecovery::recover(startupOpCtx.get())); } } else if (serverGlobalParams.clusterRole == ClusterRole::ConfigServer) { uassertStatusOK( initializeGlobalShardingStateForMongod(startupOpCtx.get(), ConnectionString::forLocal(), kDistLockProcessIdForConfigServer)); Balancer::create(startupOpCtx->getServiceContext()); } repl::getGlobalReplicationCoordinator()->startup(startupOpCtx.get()); const unsigned long long missingRepl = checkIfReplMissingFromCommandLine(startupOpCtx.get()); if (missingRepl) { log() << startupWarningsLog; log() << "** WARNING: mongod started without --replSet yet " << missingRepl << " documents are present in local.system.replset" << startupWarningsLog; log() << "** Restart with --replSet unless you are doing maintenance and " << " no other clients are connected." << startupWarningsLog; log() << "** The TTL collection monitor will not start because of this." << startupWarningsLog; log() << "** "; log() << " For more info see http://dochub.mongodb.org/core/ttlcollections"; log() << startupWarningsLog; } else { startTTLBackgroundJob(); } if (!replSettings.usingReplSets() && !replSettings.isSlave() && storageGlobalParams.engine != "devnull") { Lock::GlobalWrite lk(startupOpCtx.get()); FeatureCompatibilityVersion::setIfCleanStartup( startupOpCtx.get(), repl::StorageInterface::get(globalServiceContext)); } if (replSettings.usingReplSets() || (!replSettings.isMaster() && replSettings.isSlave()) || !internalValidateFeaturesAsMaster) { serverGlobalParams.featureCompatibility.validateFeaturesAsMaster.store(false); } } startClientCursorMonitor(); PeriodicTask::startRunningPeriodicTasks(); // MessageServer::run will return when exit code closes its socket and we don't need the // operation context anymore startupOpCtx.reset(); auto start = globalServiceContext->addAndStartTransportLayer(std::move(transportLayer)); if (!start.isOK()) { error() << "Failed to start the listener: " << start.toString(); return EXIT_NET_ERROR; } #ifndef _WIN32 mongo::signalForkSuccess(); #else if (ntservice::shouldStartService()) { ntservice::reportStatus(SERVICE_RUNNING); log() << "Service running"; } #endif if (MONGO_FAIL_POINT(shutdownAtStartup)) { log() << "starting clean exit via failpoint"; exitCleanly(EXIT_CLEAN); } MONGO_IDLE_THREAD_BLOCK; return waitForShutdown(); } ExitCode initAndListen(int listenPort) { try { return _initAndListen(listenPort); } catch (DBException& e) { log() << "exception in initAndListen: " << e.toString() << ", terminating"; return EXIT_UNCAUGHT; } catch (std::exception& e) { log() << "exception in initAndListen std::exception: " << e.what() << ", terminating"; return EXIT_UNCAUGHT; } catch (int& n) { log() << "exception in initAndListen int: " << n << ", terminating"; return EXIT_UNCAUGHT; } catch (...) { log() << "exception in initAndListen, terminating"; return EXIT_UNCAUGHT; } } } // namespace #if defined(_WIN32) ExitCode initService() { return initAndListen(serverGlobalParams.port); } #endif } // namespace mongo using namespace mongo; static int mongoDbMain(int argc, char* argv[], char** envp); #if defined(_WIN32) // In Windows, wmain() is an alternate entry point for main(), and receives the same parameters // as main() but encoded in Windows Unicode (UTF-16); "wide" 16-bit wchar_t characters. The // WindowsCommandLine object converts these wide character strings to a UTF-8 coded equivalent // and makes them available through the argv() and envp() members. This enables mongoDbMain() // to process UTF-8 encoded arguments and environment variables without regard to platform. int wmain(int argc, wchar_t* argvW[], wchar_t* envpW[]) { WindowsCommandLine wcl(argc, argvW, envpW); int exitCode = mongoDbMain(argc, wcl.argv(), wcl.envp()); quickExit(exitCode); } #else int main(int argc, char* argv[], char** envp) { int exitCode = mongoDbMain(argc, argv, envp); quickExit(exitCode); } #endif namespace { MONGO_INITIALIZER_GENERAL(ForkServer, ("EndStartupOptionHandling"), ("default")) (InitializerContext* context) { mongo::forkServerOrDie(); return Status::OK(); } } // namespace /* * This function should contain the startup "actions" that we take based on the startup config. It * is intended to separate the actions from "storage" and "validation" of our startup configuration. */ static void startupConfigActions(const std::vector& args) { // The "command" option is deprecated. For backward compatibility, still support the "run" // and "dbppath" command. The "run" command is the same as just running mongod, so just // falls through. if (moe::startupOptionsParsed.count("command")) { vector command = moe::startupOptionsParsed["command"].as>(); if (command[0].compare("dbpath") == 0) { cout << storageGlobalParams.dbpath << endl; quickExit(EXIT_SUCCESS); } if (command[0].compare("run") != 0) { cout << "Invalid command: " << command[0] << endl; printMongodHelp(moe::startupOptions); quickExit(EXIT_FAILURE); } if (command.size() > 1) { cout << "Too many parameters to 'run' command" << endl; printMongodHelp(moe::startupOptions); quickExit(EXIT_FAILURE); } } #ifdef _WIN32 ntservice::configureService(initService, moe::startupOptionsParsed, defaultServiceStrings, std::vector(), args); #endif // _WIN32 #ifdef __linux__ if (moe::startupOptionsParsed.count("shutdown") && moe::startupOptionsParsed["shutdown"].as() == true) { bool failed = false; string name = (boost::filesystem::path(storageGlobalParams.dbpath) / "mongod.lock").string(); if (!boost::filesystem::exists(name) || boost::filesystem::file_size(name) == 0) failed = true; pid_t pid; string procPath; if (!failed) { try { std::ifstream f(name.c_str()); f >> pid; procPath = (str::stream() << "/proc/" << pid); if (!boost::filesystem::exists(procPath)) failed = true; } catch (const std::exception& e) { cerr << "Error reading pid from lock file [" << name << "]: " << e.what() << endl; failed = true; } } if (failed) { std::cerr << "There doesn't seem to be a server running with dbpath: " << storageGlobalParams.dbpath << std::endl; quickExit(EXIT_FAILURE); } cout << "killing process with pid: " << pid << endl; int ret = kill(pid, SIGTERM); if (ret) { int e = errno; cerr << "failed to kill process: " << errnoWithDescription(e) << endl; quickExit(EXIT_FAILURE); } while (boost::filesystem::exists(procPath)) { sleepsecs(1); } quickExit(EXIT_SUCCESS); } #endif } MONGO_INITIALIZER_WITH_PREREQUISITES(CreateReplicationManager, ("SetGlobalEnvironment", "SSLManager", "default")) (InitializerContext* context) { auto serviceContext = getGlobalServiceContext(); repl::StorageInterface::set(serviceContext, stdx::make_unique()); auto storageInterface = repl::StorageInterface::get(serviceContext); repl::ReplicationProcess::set(serviceContext, stdx::make_unique(storageInterface)); auto replicationProcess = repl::ReplicationProcess::get(serviceContext); repl::TopologyCoordinatorImpl::Options topoCoordOptions; topoCoordOptions.maxSyncSourceLagSecs = Seconds(repl::maxSyncSourceLagSecs); topoCoordOptions.clusterRole = serverGlobalParams.clusterRole; auto logicalClock = stdx::make_unique(serviceContext); LogicalClock::set(serviceContext, std::move(logicalClock)); auto hookList = stdx::make_unique(); hookList->addHook(stdx::make_unique(serviceContext)); auto replCoord = stdx::make_unique( serviceContext, getGlobalReplSettings(), stdx::make_unique(serviceContext, storageInterface), executor::makeNetworkInterface( "NetworkInterfaceASIO-Replication", nullptr, std::move(hookList)), stdx::make_unique(topoCoordOptions), replicationProcess, storageInterface, static_cast(curTimeMillis64())); repl::ReplicationCoordinator::set(serviceContext, std::move(replCoord)); repl::setOplogCollectionName(); return Status::OK(); } #ifdef MONGO_CONFIG_SSL MONGO_INITIALIZER_GENERAL(setSSLManagerType, MONGO_NO_PREREQUISITES, ("SSLManager")) (InitializerContext* context) { isSSLServer = true; return Status::OK(); } #endif #if !defined(__has_feature) #define __has_feature(x) 0 #endif // NOTE: This function may be called at any time after // registerShutdownTask is called below. It must not depend on the // prior execution of mongo initializers or the existence of threads. static void shutdownTask() { auto serviceContext = getGlobalServiceContext(); Client::initThreadIfNotAlready(); Client& client = cc(); ServiceContext::UniqueOperationContext uniqueTxn; OperationContext* opCtx = client.getOperationContext(); if (!opCtx && serviceContext->getGlobalStorageEngine()) { uniqueTxn = client.makeOperationContext(); opCtx = uniqueTxn.get(); } log(LogComponent::kNetwork) << "shutdown: going to close listening sockets..." << endl; ListeningSockets::get()->closeAll(); log(LogComponent::kNetwork) << "shutdown: going to flush diaglog..." << endl; _diaglog.flush(); if (opCtx) { // This can wait a long time while we drain the secondary's apply queue, especially if it is // building an index. repl::ReplicationCoordinator::get(opCtx)->shutdown(opCtx); } if (serviceContext) serviceContext->setKillAllOperations(); ReplicaSetMonitor::shutdown(); if (auto sr = grid.shardRegistry()) { // TODO: race: sr is a naked pointer sr->shutdown(); } #if __has_feature(address_sanitizer) if (auto sep = checked_cast(serviceContext->getServiceEntryPoint())) { // When running under address sanitizer, we get false positive leaks due to disorder around // the lifecycle of a connection and request. When we are running under ASAN, we try a lot // harder to dry up the server from active connections before going on to really shut down. log(LogComponent::kNetwork) << "shutdown: going to close all sockets because ASAN is active..."; getGlobalServiceContext()->getTransportLayer()->shutdown(); // Close all sockets in a detached thread, and then wait for the number of active // connections to reach zero. Give the detached background thread a 10 second deadline. If // we haven't closed drained all active operations within that deadline, just keep going // with shutdown: the OS will do it for us when the process terminates. stdx::packaged_task dryOutTask([sep] { // There isn't currently a way to wait on the TicketHolder to have all its tickets back, // unfortunately. So, busy wait in this detached thread. while (true) { const auto runningWorkers = sep->getNumberOfActiveWorkerThreads(); if (runningWorkers == 0) { log(LogComponent::kNetwork) << "shutdown: no running workers found..."; break; } log(LogComponent::kNetwork) << "shutdown: still waiting on " << runningWorkers << " active workers to drain... "; mongo::sleepFor(Milliseconds(250)); } }); auto dryNotification = dryOutTask.get_future(); stdx::thread(std::move(dryOutTask)).detach(); if (dryNotification.wait_for(Seconds(10).toSystemDuration()) != stdx::future_status::ready) { log(LogComponent::kNetwork) << "shutdown: exhausted grace period for" << " active workers to drain; continuing with shutdown... "; } } #endif // Shutdown Full-Time Data Capture stopFTDC(); if (opCtx) { ShardingState::get(opCtx)->shutDown(opCtx); } // We should always be able to acquire the global lock at shutdown. // // TODO: This call chain uses the locker directly, because we do not want to start an // operation context, which also instantiates a recovery unit. Also, using the // lockGlobalBegin/lockGlobalComplete sequence, we avoid taking the flush lock. // // For a Windows service, dbexit does not call exit(), so we must leak the lock outside // of this function to prevent any operations from running that need a lock. // DefaultLockerImpl* globalLocker = new DefaultLockerImpl(); LockResult result = globalLocker->lockGlobalBegin(MODE_X, Milliseconds::max()); if (result == LOCK_WAITING) { result = globalLocker->lockGlobalComplete(Milliseconds::max()); } invariant(LOCK_OK == result); // Global storage engine may not be started in all cases before we exit if (serviceContext && serviceContext->getGlobalStorageEngine()) { serviceContext->shutdownGlobalStorageEngineCleanly(); } // We drop the scope cache because leak sanitizer can't see across the // thread we use for proxying MozJS requests. Dropping the cache cleans up // the memory and makes leak sanitizer happy. ScriptEngine::dropScopeCache(); log(LogComponent::kControl) << "now exiting" << endl; audit::logShutdown(&cc()); } static int mongoDbMain(int argc, char* argv[], char** envp) { registerShutdownTask(shutdownTask); setupSignalHandlers(); srand(static_cast(curTimeMicros64())); Status status = mongo::runGlobalInitializers(argc, argv, envp); if (!status.isOK()) { severe(LogComponent::kControl) << "Failed global initialization: " << status; quickExit(EXIT_FAILURE); } startupConfigActions(std::vector(argv, argv + argc)); cmdline_utils::censorArgvArray(argc, argv); if (!initializeServerGlobalState()) quickExit(EXIT_FAILURE); // Per SERVER-7434, startSignalProcessingThread() must run after any forks // (initializeServerGlobalState()) and before creation of any other threads. startSignalProcessingThread(); #if defined(_WIN32) if (ntservice::shouldStartService()) { ntservice::startService(); // exits directly and so never reaches here either. } #endif StartupTest::runTests(); ExitCode exitCode = initAndListen(serverGlobalParams.port); exitCleanly(exitCode); return 0; }