bool ReplicationCoordinatorExternalStateImpl::isReadConcernSnapshotSupportedByStorageEngine(
OperationContext* opCtx) const {
auto storageEngine = opCtx->getServiceContext()->getGlobalStorageEngine();
// This should never be called if the storage engine has not been initialized.
invariant(storageEngine);
return storageEngine->supportsReadConcernSnapshot();
}
ExitCode _initAndListen(int listenPort) {
Client::initThread("initandlisten");
initWireSpec();
auto serviceContext = getGlobalServiceContext();
serviceContext->setFastClockSource(FastClockSourceFactory::create(Milliseconds(10)));
auto opObserverRegistry = stdx::make_unique<OpObserverRegistry>();
opObserverRegistry->addObserver(stdx::make_unique<OpObserverShardingImpl>());
opObserverRegistry->addObserver(stdx::make_unique<UUIDCatalogObserver>());
if (serverGlobalParams.clusterRole == ClusterRole::ShardServer) {
opObserverRegistry->addObserver(stdx::make_unique<ShardServerOpObserver>());
} else if (serverGlobalParams.clusterRole == ClusterRole::ConfigServer) {
opObserverRegistry->addObserver(stdx::make_unique<ConfigServerOpObserver>());
}
setupFreeMonitoringOpObserver(opObserverRegistry.get());
serviceContext->setOpObserver(std::move(opObserverRegistry));
DBDirectClientFactory::get(serviceContext).registerImplementation([](OperationContext* opCtx) {
return std::unique_ptr<DBClientBase>(new DBDirectClient(opCtx));
});
const repl::ReplSettings& replSettings =
repl::ReplicationCoordinator::get(serviceContext)->getSettings();
{
ProcessId pid = ProcessId::getCurrent();
LogstreamBuilder l = log(LogComponent::kControl);
l << "MongoDB starting : pid=" << pid << " port=" << serverGlobalParams.port
<< " dbpath=" << storageGlobalParams.dbpath;
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();
serviceContext->setServiceEntryPoint(
stdx::make_unique<ServiceEntryPointMongod>(serviceContext));
if (!storageGlobalParams.repair) {
auto tl =
transport::TransportLayerManager::createWithConfig(&serverGlobalParams, serviceContext);
auto res = tl->setup();
if (!res.isOK()) {
error() << "Failed to set up listener: " << res;
return EXIT_NET_ERROR;
}
serviceContext->setTransportLayer(std::move(tl));
}
// Set up the periodic runner for background job execution. This is required to be running
// before the storage engine is initialized.
auto runner = makePeriodicRunner(serviceContext);
runner->startup();
serviceContext->setPeriodicRunner(std::move(runner));
initializeStorageEngine(serviceContext, StorageEngineInitFlags::kNone);
#ifdef MONGO_CONFIG_WIREDTIGER_ENABLED
if (EncryptionHooks::get(serviceContext)->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());
for (auto&& e : storageElement.Obj()) {
// 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 (isRegisteredStorageEngine(serviceContext, e.fieldName())) {
warning() << "Detected configuration for non-active storage engine "
<< e.fieldName() << " when current storage engine is "
<< storageGlobalParams.engine;
}
}
}
// Disallow running a storage engine that doesn't support capped collections with --profile
if (!serviceContext->getStorageEngine()->supportsCappedCollections() &&
serverGlobalParams.defaultProfile != 0) {
log() << "Running " << storageGlobalParams.engine << " with profiling is not supported. "
<< "Make sure you are not using --profile.";
exitCleanly(EXIT_BADOPTIONS);
}
// Disallow running WiredTiger with --nojournal in a replica set
if (storageGlobalParams.engine == "wiredTiger" && !storageGlobalParams.dur &&
replSettings.usingReplSets()) {
log() << "Running wiredTiger without journaling in a replica set is not "
<< "supported. Make sure you are not using --nojournal and that "
<< "storage.journal.enabled is not set to 'false'.";
exitCleanly(EXIT_BADOPTIONS);
}
logMongodStartupWarnings(storageGlobalParams, serverGlobalParams, serviceContext);
#ifdef MONGO_CONFIG_SSL
if (sslGlobalParams.sslAllowInvalidCertificates &&
((serverGlobalParams.clusterAuthMode.load() == ServerGlobalParams::ClusterAuthMode_x509) ||
sequenceContains(saslGlobalParams.authenticationMechanisms, "MONGODB-X509"))) {
log() << "** WARNING: While invalid X509 certificates may be used to" << startupWarningsLog;
log() << "** connect to this server, they will not be considered"
<< startupWarningsLog;
log() << "** permissible for authentication." << startupWarningsLog;
log() << startupWarningsLog;
}
#endif
{
std::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));
}
initializeSNMP();
if (!storageGlobalParams.readOnly) {
boost::filesystem::remove_all(storageGlobalParams.dbpath + "/_tmp/");
}
if (mongodGlobalParams.scriptingEnabled) {
ScriptEngine::setup();
}
auto startupOpCtx = serviceContext->makeOperationContext(&cc());
bool canCallFCVSetIfCleanStartup =
!storageGlobalParams.readOnly && (storageGlobalParams.engine != "devnull");
if (canCallFCVSetIfCleanStartup && !replSettings.usingReplSets()) {
Lock::GlobalWrite lk(startupOpCtx.get());
FeatureCompatibilityVersion::setIfCleanStartup(startupOpCtx.get(),
repl::StorageInterface::get(serviceContext));
}
auto swNonLocalDatabases = repairDatabasesAndCheckVersion(startupOpCtx.get());
if (!swNonLocalDatabases.isOK()) {
// SERVER-31611 introduced a return value to `repairDatabasesAndCheckVersion`. Previously,
// a failing condition would fassert. SERVER-31611 covers a case where the binary (3.6) is
// refusing to start up because it refuses acknowledgement of FCV 3.2 and requires the
// user to start up with an older binary. Thus shutting down the server must leave the
// datafiles in a state that the older binary can start up. This requires going through a
// clean shutdown.
//
// The invariant is *not* a statement that `repairDatabasesAndCheckVersion` must return
// `MustDowngrade`. Instead, it is meant as a guardrail to protect future developers from
// accidentally buying into this behavior. New errors that are returned from the method
// may or may not want to go through a clean shutdown, and they likely won't want the
// program to return an exit code of `EXIT_NEED_DOWNGRADE`.
severe(LogComponent::kControl) << "** IMPORTANT: "
<< swNonLocalDatabases.getStatus().reason();
invariant(swNonLocalDatabases == ErrorCodes::MustDowngrade);
exitCleanly(EXIT_NEED_DOWNGRADE);
}
// Assert that the in-memory featureCompatibilityVersion parameter has been explicitly set. If
// we are part of a replica set and are started up with no data files, we do not set the
// featureCompatibilityVersion until a primary is chosen. For this case, we expect the in-memory
// featureCompatibilityVersion parameter to still be uninitialized until after startup.
if (canCallFCVSetIfCleanStartup &&
(!replSettings.usingReplSets() || swNonLocalDatabases.getValue())) {
invariant(serverGlobalParams.featureCompatibility.isVersionInitialized());
}
if (storageGlobalParams.upgrade) {
log() << "finished checking dbs";
exitCleanly(EXIT_CLEAN);
}
// Start up health log writer thread.
HealthLog::get(startupOpCtx.get()).startup();
auto const globalAuthzManager = AuthorizationManager::get(serviceContext);
uassertStatusOK(globalAuthzManager->initialize(startupOpCtx.get()));
// This is for security on certain platforms (nonce generation)
srand((unsigned)(curTimeMicros64()) ^ (unsigned(uintptr_t(&startupOpCtx))));
if (globalAuthzManager->shouldValidateAuthSchemaOnStartup()) {
Status status = verifySystemIndexes(startupOpCtx.get());
if (!status.isOK()) {
log() << redact(status);
if (status == ErrorCodes::AuthSchemaIncompatible) {
exitCleanly(EXIT_NEED_UPGRADE);
} else if (status == ErrorCodes::NotMaster) {
// Try creating the indexes if we become master. If we do not become master,
// the master will create the indexes and we will replicate them.
} else {
quickExit(EXIT_FAILURE);
}
}
// 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;
log() << "To manually repair the 'authSchema' document in the admin.system.version "
"collection, start up with --setParameter "
"startupAuthSchemaValidation=false to disable validation.";
exitCleanly(EXIT_NEED_UPGRADE);
}
if (foundSchemaVersion <= AuthorizationManager::schemaVersion26Final) {
log() << "This server is using MONGODB-CR, an authentication mechanism which "
<< "has been removed from MongoDB 4.0. In order to upgrade the auth schema, "
<< "first downgrade MongoDB binaries to version 3.6 and then run the "
<< "authSchemaUpgrade command. "
<< "See http://dochub.mongodb.org/core/3.0-upgrade-to-scram-sha-1";
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;
}
// This function may take the global lock.
auto shardingInitialized = ShardingInitializationMongoD::get(startupOpCtx.get())
->initializeShardingAwarenessIfNeeded(startupOpCtx.get());
if (shardingInitialized) {
waitForShardRegistryReload(startupOpCtx.get()).transitional_ignore();
}
auto storageEngine = serviceContext->getStorageEngine();
invariant(storageEngine);
BackupCursorHooks::initialize(serviceContext, storageEngine);
if (!storageGlobalParams.readOnly) {
if (storageEngine->supportsCappedCollections()) {
logStartup(startupOpCtx.get());
}
startMongoDFTDC();
startFreeMonitoring(serviceContext);
restartInProgressIndexesFromLastShutdown(startupOpCtx.get());
if (serverGlobalParams.clusterRole == ClusterRole::ShardServer) {
// Note: For replica sets, ShardingStateRecovery happens on transition to primary.
if (!repl::ReplicationCoordinator::get(startupOpCtx.get())->isReplEnabled()) {
if (ShardingState::get(startupOpCtx.get())->enabled()) {
uassertStatusOK(ShardingStateRecovery::recover(startupOpCtx.get()));
}
}
} else if (serverGlobalParams.clusterRole == ClusterRole::ConfigServer) {
initializeGlobalShardingStateForMongoD(startupOpCtx.get(),
ConnectionString::forLocal(),
kDistLockProcessIdForConfigServer);
Balancer::create(startupOpCtx->getServiceContext());
ShardingCatalogManager::create(
startupOpCtx->getServiceContext(),
makeShardingTaskExecutor(executor::makeNetworkInterface("AddShard-TaskExecutor")));
Grid::get(startupOpCtx.get())->setShardingInitialized();
} else if (replSettings.usingReplSets()) { // standalone replica set
auto keysCollectionClient = stdx::make_unique<KeysCollectionClientDirect>();
auto keyManager = std::make_shared<KeysCollectionManager>(
KeysCollectionManager::kKeyManagerPurposeString,
std::move(keysCollectionClient),
Seconds(KeysRotationIntervalSec));
keyManager->startMonitoring(startupOpCtx->getServiceContext());
LogicalTimeValidator::set(startupOpCtx->getServiceContext(),
stdx::make_unique<LogicalTimeValidator>(keyManager));
}
repl::ReplicationCoordinator::get(startupOpCtx.get())->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() << "** Database contents may appear inconsistent with the oplog and may "
"appear to not contain"
<< startupWarningsLog;
log() << "** writes that were visible when this node was running as part of a "
"replica set."
<< 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() || !internalValidateFeaturesAsMaster) {
serverGlobalParams.validateFeaturesAsMaster.store(false);
}
}
startClientCursorMonitor();
PeriodicTask::startRunningPeriodicTasks();
SessionKiller::set(serviceContext,
std::make_shared<SessionKiller>(serviceContext, killSessionsLocal));
// Start up a background task to periodically check for and kill expired transactions; and a
// background task to periodically check for and decrease cache pressure by decreasing the
// target size setting for the storage engine's window of available snapshots.
//
// Only do this on storage engines supporting snapshot reads, which hold resources we wish to
// release periodically in order to avoid storage cache pressure build up.
if (storageEngine->supportsReadConcernSnapshot()) {
startPeriodicThreadToAbortExpiredTransactions(serviceContext);
startPeriodicThreadToDecreaseSnapshotHistoryCachePressure(serviceContext);
}
// Set up the logical session cache
LogicalSessionCacheServer kind = LogicalSessionCacheServer::kStandalone;
if (serverGlobalParams.clusterRole == ClusterRole::ShardServer) {
kind = LogicalSessionCacheServer::kSharded;
} else if (serverGlobalParams.clusterRole == ClusterRole::ConfigServer) {
kind = LogicalSessionCacheServer::kConfigServer;
} else if (replSettings.usingReplSets()) {
kind = LogicalSessionCacheServer::kReplicaSet;
}
auto sessionCache = makeLogicalSessionCacheD(kind);
LogicalSessionCache::set(serviceContext, std::move(sessionCache));
// MessageServer::run will return when exit code closes its socket and we don't need the
// operation context anymore
startupOpCtx.reset();
auto start = serviceContext->getServiceExecutor()->start();
if (!start.isOK()) {
error() << "Failed to start the service executor: " << start;
return EXIT_NET_ERROR;
}
start = serviceContext->getServiceEntryPoint()->start();
if (!start.isOK()) {
error() << "Failed to start the service entry point: " << start;
return EXIT_NET_ERROR;
}
if (!storageGlobalParams.repair) {
start = serviceContext->getTransportLayer()->start();
if (!start.isOK()) {
error() << "Failed to start the listener: " << start.toString();
return EXIT_NET_ERROR;
}
}
serviceContext->notifyStartupComplete();
#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();
}