diff options
19 files changed, 464 insertions, 8 deletions
diff --git a/buildscripts/resmokeconfig/suites/sharding_last_stable_mongos_and_mixed_shards.yml b/buildscripts/resmokeconfig/suites/sharding_last_stable_mongos_and_mixed_shards.yml index 9aac8d16b57..fab6f7fda0c 100644 --- a/buildscripts/resmokeconfig/suites/sharding_last_stable_mongos_and_mixed_shards.yml +++ b/buildscripts/resmokeconfig/suites/sharding_last_stable_mongos_and_mixed_shards.yml @@ -184,6 +184,8 @@ selector: - jstests/sharding/mongos_precache_routing_info.js - jstests/sharding/warm_up_connection_pool.js - jstests/sharding/collation_shard_targeting_hashed_shard_key.js + # Enable when SERVER-62065 is released in 4.0 + - jstests/sharding/repair_sharded_collection_history.js exclude_with_any_tags: - multiversion_incompatible diff --git a/jstests/core/views/views_all_commands.js b/jstests/core/views/views_all_commands.js index f8d1be30b7b..345c408e321 100644 --- a/jstests/core/views/views_all_commands.js +++ b/jstests/core/views/views_all_commands.js @@ -91,6 +91,7 @@ let viewsCommandTests = { _configsvrMovePrimary: {skip: isAnInternalCommand}, _configsvrRemoveShard: {skip: isAnInternalCommand}, _configsvrRemoveShardFromZone: {skip: isAnInternalCommand}, + _configsvrRepairShardedCollectionChunksHistory: {skip: isAnInternalCommand}, _configsvrShardCollection: {skip: isAnInternalCommand}, _configsvrUpdateZoneKeyRange: {skip: isAnInternalCommand}, _cpuProfilerStart: {skip: isAnInternalCommand}, @@ -428,6 +429,13 @@ let viewsCommandTests = { ], repairCursor: {command: {repairCursor: "view"}, expectFailure: true}, repairDatabase: {skip: isUnrelated}, + repairShardedCollectionChunksHistory: { + command: {repairShardedCollectionChunksHistory: "test.view"}, + skipStandalone: true, + isAdminCommand: true, + expectFailure: true, + expectedErrorCode: ErrorCodes.NamespaceNotFound, + }, replSetAbortPrimaryCatchUp: {skip: isUnrelated}, replSetFreeze: {skip: isUnrelated}, replSetGetConfig: {skip: isUnrelated}, diff --git a/jstests/replsets/db_reads_while_recovering_all_commands.js b/jstests/replsets/db_reads_while_recovering_all_commands.js index 05dade63ca3..37ed95f1015 100644 --- a/jstests/replsets/db_reads_while_recovering_all_commands.js +++ b/jstests/replsets/db_reads_while_recovering_all_commands.js @@ -49,6 +49,7 @@ const allCommands = { _configsvrRefineCollectionShardKey: {skip: isPrimaryOnly}, _configsvrRemoveShard: {skip: isPrimaryOnly}, _configsvrRemoveShardFromZone: {skip: isPrimaryOnly}, + _configsvrRepairShardedCollectionChunksHistory: {skip: isAnInternalCommand}, _configsvrShardCollection: {skip: isPrimaryOnly}, _configsvrUpdateZoneKeyRange: {skip: isPrimaryOnly}, _flushDatabaseCacheUpdates: {skip: isPrimaryOnly}, @@ -251,6 +252,7 @@ const allCommands = { reIndex: {skip: isNotAUserDataRead}, renameCollection: {skip: isPrimaryOnly}, repairDatabase: {skip: isNotAUserDataRead}, + repairShardedCollectionChunksHistory: {skip: isPrimaryOnly}, repairCursor: {skip: isNotAUserDataRead}, replSetAbortPrimaryCatchUp: {skip: isNotAUserDataRead}, replSetFreeze: {skip: isNotAUserDataRead}, diff --git a/jstests/sharding/database_and_shard_versioning_all_commands.js b/jstests/sharding/database_and_shard_versioning_all_commands.js index 52a5bd35fe7..ca9d56c8a80 100644 --- a/jstests/sharding/database_and_shard_versioning_all_commands.js +++ b/jstests/sharding/database_and_shard_versioning_all_commands.js @@ -376,6 +376,7 @@ let testCases = { assert(mongosConn.getDB(dbName).getCollection(collName + "_renamed").drop()); } }, + repairShardedCollectionChunksHistory: {skip: "always targets the config server"}, replSetGetStatus: {skip: "not supported in mongos"}, resetError: {skip: "not on a user database"}, restartCatalog: {skip: "not on a user database"}, diff --git a/jstests/sharding/repair_sharded_collection_history.js b/jstests/sharding/repair_sharded_collection_history.js new file mode 100644 index 00000000000..e069873d1cd --- /dev/null +++ b/jstests/sharding/repair_sharded_collection_history.js @@ -0,0 +1,59 @@ +(function() { +"use strict"; + +load("jstests/libs/feature_compatibility_version.js"); + +var st = new ShardingTest({ + shards: 1, +}); + +let configPrimary = st.configRS.getPrimary(); +let configPrimaryAdminDB = configPrimary.getDB("admin"); +let shardPrimary = st.rs0.getPrimary(); +let shardPrimaryAdminDB = shardPrimary.getDB("admin"); +let shardPrimaryConfigDB = shardPrimary.getDB("config"); + +let testDB = st.s.getDB("test1"); + +// Create a sharded collection with primary shard 0. +assert.commandWorked(st.s.adminCommand({enableSharding: testDB.getName()})); +st.ensurePrimaryShard(testDB.getName(), st.shard0.shardName); +assert.commandWorked(st.s.adminCommand({shardCollection: testDB.foo.getFullName(), key: {a: 1}})); +assert.commandWorked(st.s.adminCommand({split: testDB.foo.getFullName(), middle: {a: 0}})); +assert.commandWorked(st.s.adminCommand({split: testDB.foo.getFullName(), middle: {a: -1000}})); +assert.commandWorked(st.s.adminCommand({split: testDB.foo.getFullName(), middle: {a: +1000}})); + +assert.writeOK(st.s.getDB("test1").foo.insert({_id: "id1", a: 1})); +assert.neq(null, st.s.getDB("test1").foo.findOne({_id: "id1", a: 1})); + +assert.writeOK(st.s.getDB("test1").foo.insert({_id: "id2", a: -1})); +assert.neq(null, st.s.getDB("test1").foo.findOne({_id: "id2", a: -1})); + +// Manually clear the 'historyIsAt40' field from the config server and the history entries from +// the shards' cache collections in order to simulate a wrong upgrade due to SERVER-62065 +assert.writeOK(st.s.getDB("config").chunks.update( + {ns: 'test1.foo'}, {'$unset': {historyIsAt40: ''}}, {multi: true})); +assert.writeOK(shardPrimaryConfigDB.getCollection("cache.chunks.test1.foo") + .update({}, {'$unset': {history: ''}}, {multi: true})); + +assert.commandWorked(st.s.adminCommand({repairShardedCollectionChunksHistory: 'test1.foo'})); + +// Make sure chunks for test1.foo were given history after repair. +var chunks = st.s.getDB("config").getCollection("chunks").find({ns: "test1.foo"}).toArray(); +assert.eq(chunks.length, 4); +chunks.forEach((chunk) => { + assert.neq(null, chunk); + assert(chunk.hasOwnProperty("history"), "test1.foo does not have a history after repair"); + assert(chunk.hasOwnProperty("historyIsAt40"), + "test1.foo does not have a historyIsAt40 after repair"); +}); +chunks = shardPrimaryConfigDB.getCollection("cache.chunks.test1.foo").find().toArray(); +assert.eq(chunks.length, 4); +chunks.forEach((chunk) => { + assert.neq(null, chunk); + assert(chunk.hasOwnProperty("history"), + "test1.foo does not have a history on the shard after repair"); +}); + +st.stop(); +})(); diff --git a/jstests/sharding/safe_secondary_reads_drop_recreate.js b/jstests/sharding/safe_secondary_reads_drop_recreate.js index 575746f83c1..425eeffc634 100644 --- a/jstests/sharding/safe_secondary_reads_drop_recreate.js +++ b/jstests/sharding/safe_secondary_reads_drop_recreate.js @@ -253,6 +253,7 @@ let testCases = { removeShardFromZone: {skip: "primary only"}, renameCollection: {skip: "primary only"}, repairCursor: {skip: "does not return user data"}, + repairShardedCollectionChunksHistory: {skip: "does not return user data"}, replSetAbortPrimaryCatchUp: {skip: "does not return user data"}, replSetFreeze: {skip: "does not return user data"}, replSetGetConfig: {skip: "does not return user data"}, diff --git a/jstests/sharding/safe_secondary_reads_single_migration_suspend_range_deletion.js b/jstests/sharding/safe_secondary_reads_single_migration_suspend_range_deletion.js index 88318a11d31..540855cc851 100644 --- a/jstests/sharding/safe_secondary_reads_single_migration_suspend_range_deletion.js +++ b/jstests/sharding/safe_secondary_reads_single_migration_suspend_range_deletion.js @@ -288,6 +288,7 @@ let testCases = { removeShardFromZone: {skip: "primary only"}, renameCollection: {skip: "primary only"}, repairCursor: {skip: "does not return user data"}, + repairShardedCollectionChunksHistory: {skip: "does not return user data"}, replSetAbortPrimaryCatchUp: {skip: "does not return user data"}, replSetFreeze: {skip: "does not return user data"}, replSetGetConfig: {skip: "does not return user data"}, diff --git a/jstests/sharding/safe_secondary_reads_single_migration_waitForDelete.js b/jstests/sharding/safe_secondary_reads_single_migration_waitForDelete.js index 9469bb43581..cda4797802f 100644 --- a/jstests/sharding/safe_secondary_reads_single_migration_waitForDelete.js +++ b/jstests/sharding/safe_secondary_reads_single_migration_waitForDelete.js @@ -258,6 +258,7 @@ let testCases = { removeShardFromZone: {skip: "primary only"}, renameCollection: {skip: "primary only"}, repairCursor: {skip: "does not return user data"}, + repairShardedCollectionChunksHistory: {skip: "does not return user data"}, replSetAbortPrimaryCatchUp: {skip: "does not return user data"}, replSetFreeze: {skip: "does not return user data"}, replSetGetConfig: {skip: "does not return user data"}, diff --git a/src/mongo/db/s/SConscript b/src/mongo/db/s/SConscript index e4418131d92..cc37c48d1d7 100644 --- a/src/mongo/db/s/SConscript +++ b/src/mongo/db/s/SConscript @@ -331,6 +331,7 @@ env.Library( 'config/configsvr_move_primary_command.cpp', 'config/configsvr_remove_shard_command.cpp', 'config/configsvr_remove_shard_from_zone_command.cpp', + 'config/configsvr_repair_sharded_collection_chunks_history_command.cpp', 'config/configsvr_shard_collection_command.cpp', 'config/configsvr_split_chunk_command.cpp', 'config/configsvr_update_zone_key_range_command.cpp', diff --git a/src/mongo/db/s/config/configsvr_repair_sharded_collection_chunks_history_command.cpp b/src/mongo/db/s/config/configsvr_repair_sharded_collection_chunks_history_command.cpp new file mode 100644 index 00000000000..77352ff3611 --- /dev/null +++ b/src/mongo/db/s/config/configsvr_repair_sharded_collection_chunks_history_command.cpp @@ -0,0 +1,153 @@ + +/** + * Copyright (C) 2021 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 + * <http://www.mongodb.com/licensing/server-side-public-license>. + * + * 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_LOG_DEFAULT_COMPONENT ::mongo::logger::LogComponent::kSharding + +#include "mongo/platform/basic.h" + +#include "mongo/db/auth/action_type.h" +#include "mongo/db/auth/authorization_session.h" +#include "mongo/db/auth/privilege.h" +#include "mongo/db/commands.h" +#include "mongo/db/logical_clock.h" +#include "mongo/db/namespace_string.h" +#include "mongo/db/s/config/sharding_catalog_manager.h" +#include "mongo/s/grid.h" +#include "mongo/util/log.h" + +namespace mongo { +namespace { + +class ConfigSvrRepairShardedCollectionChunksHistoryCommand : public BasicCommand { +public: + ConfigSvrRepairShardedCollectionChunksHistoryCommand() + : BasicCommand("_configsvrRepairShardedCollectionChunksHistory") {} + + std::string help() const override { + return "Internal command, which is exported by the sharding config server. Do not call " + "directly."; + } + + AllowedOnSecondary secondaryAllowed(ServiceContext*) const override { + return AllowedOnSecondary::kNever; + } + + bool adminOnly() const override { + return true; + } + + bool supportsWriteConcern(const BSONObj& cmd) const override { + return true; + } + + std::string parseNs(const std::string& unusedDbName, const BSONObj& cmdObj) const override { + return CommandHelpers::parseNsFullyQualified(cmdObj); + } + + Status checkAuthForCommand(Client* client, + const std::string& dbname, + const BSONObj& cmdObj) const override { + if (!AuthorizationSession::get(client)->isAuthorizedForActionsOnResource( + ResourcePattern::forClusterResource(), ActionType::internal)) { + return Status(ErrorCodes::Unauthorized, "Unauthorized"); + } + return Status::OK(); + } + + bool run(OperationContext* opCtx, + const std::string& unusedDbName, + const BSONObj& cmdObj, + BSONObjBuilder& result) override { + uassert(ErrorCodes::IllegalOperation, + "_configsvrRepairShardedCollectionChunksHistory can only be run on config servers", + serverGlobalParams.clusterRole == ClusterRole::ConfigServer); + + // Set the operation context read concern level to local for reads into the config database. + repl::ReadConcernArgs::get(opCtx) = + repl::ReadConcernArgs(repl::ReadConcernLevel::kLocalReadConcern); + + uassert(ErrorCodes::InvalidOptions, + str::stream() << "_configsvrRepairShardedCollectionChunksHistory must be called " + "with majority writeConcern, got " + << cmdObj, + opCtx->getWriteConcern().wMode == WriteConcernOptions::kMajority); + + const NamespaceString nss{parseNs(unusedDbName, cmdObj)}; + + auto const catalogClient = Grid::get(opCtx)->catalogClient(); + auto collection = + uassertStatusOK( + catalogClient->getCollection(opCtx, nss, repl::ReadConcernLevel::kLocalReadConcern)) + .value; + + if (cmdObj["force"].booleanSafe()) { + LOG(0) << "Resetting the 'historyIsAt40' field for all chunks in collection " + << nss.ns() << " in order to force all chunks' history to get recreated"; + + BatchedCommandRequest request([&] { + write_ops::Update updateOp(ChunkType::ConfigNS); + updateOp.setUpdates({[&] { + write_ops::UpdateOpEntry entry; + entry.setQ(BSON("ns" << nss.ns())); + entry.setU(BSON("$unset" << BSON(ChunkType::historyIsAt40() << ""))); + entry.setUpsert(false); + entry.setMulti(true); + return entry; + }()}); + return updateOp; + }()); + request.setWriteConcern(ShardingCatalogClient::kLocalWriteConcern.toBSON()); + + const auto configShard = Grid::get(opCtx)->shardRegistry()->getConfigShard(); + auto response = configShard->runBatchWriteCommand(opCtx, + Shard::kDefaultConfigCommandTimeout, + request, + Shard::RetryPolicy::kIdempotent); + uassertStatusOK(response.toStatus()); + + uassert(ErrorCodes::Error(5760502), + str::stream() << "No chunks found for collection " << nss.ns(), + response.getN() > 0); + } + + auto validAfter = LogicalClock::get(opCtx)->getClusterTime().asTimestamp(); + + ShardingCatalogManager::get(opCtx)->upgradeChunksHistory( + opCtx, nss, collection.getEpoch(), validAfter); + + Grid::get(opCtx)->catalogCache()->invalidateShardedCollection(nss); + + return true; + } + +} configSvrRepairShardedCollectionChunksHistoryCommand; + +} // namespace +} // namespace mongo diff --git a/src/mongo/db/s/config/sharding_catalog_manager.h b/src/mongo/db/s/config/sharding_catalog_manager.h index 11131fdfdaf..c0fa607182b 100644 --- a/src/mongo/db/s/config/sharding_catalog_manager.h +++ b/src/mongo/db/s/config/sharding_catalog_manager.h @@ -430,6 +430,18 @@ public: Lock::ExclusiveLock lockZoneMutex(OperationContext* opCtx); + // + // Upgrade/downgrade + // + + /** + * Upgrade the chunk metadata to include the history field. + */ + void upgradeChunksHistory(OperationContext* opCtx, + const NamespaceString& nss, + const OID& collectionEpoch, + const Timestamp validAfter); + private: /** * Performs the necessary checks for version compatibility and creates a new config.version 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 d821e141506..1856f3dec14 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 @@ -42,8 +42,10 @@ #include "mongo/db/dbdirectclient.h" #include "mongo/db/namespace_string.h" #include "mongo/db/operation_context.h" +#include "mongo/db/repl/repl_client_info.h" #include "mongo/db/s/sharding_logging.h" #include "mongo/db/s/sharding_runtime_d_params_gen.h" +#include "mongo/db/write_concern.h" #include "mongo/rpc/get_status_from_command_result.h" #include "mongo/s/catalog/sharding_catalog_client.h" #include "mongo/s/catalog/type_chunk.h" @@ -1111,6 +1113,94 @@ StatusWith<ChunkType> ShardingCatalogManager::_findChunkOnConfig(OperationContex return ChunkType::fromConfigBSON(origChunks.front()); } +void ShardingCatalogManager::upgradeChunksHistory(OperationContext* opCtx, + const NamespaceString& nss, + const OID& collectionEpoch, + const Timestamp validAfter) { + auto const catalogClient = Grid::get(opCtx)->catalogClient(); + const auto shardRegistry = Grid::get(opCtx)->shardRegistry(); + + // Take _kChunkOpLock in exclusive mode to prevent concurrent chunk splits, merges, and + // migrations. + Lock::ExclusiveLock lk(opCtx->lockState(), _kChunkOpLock); + + auto const configShard = shardRegistry->getConfigShard(); + const auto chunksVector = [&] { + auto findChunksResponse = uassertStatusOK( + configShard->exhaustiveFindOnConfig(opCtx, + ReadPreferenceSetting{ReadPreference::PrimaryOnly}, + repl::ReadConcernLevel::kLocalReadConcern, + ChunkType::ConfigNS, + BSON("ns" << nss.ns()), + BSONObj(), + boost::none)); + uassert(ErrorCodes::Error(5760503), + str::stream() << "No chunks found for collection " << nss.ns(), + !findChunksResponse.docs.empty()); + return std::move(findChunksResponse.docs); + }(); + + const auto currentCollectionVersion = + uassertStatusOK(_findCollectionVersion(opCtx, nss, collectionEpoch)); + + // Bump the major version in order to be guaranteed to trigger refresh on every shard + ChunkVersion newCollectionVersion( + currentCollectionVersion.majorVersion() + 1, 0, currentCollectionVersion.epoch()); + std::set<ShardId> changedShardIds; + for (const auto& chunk : chunksVector) { + auto upgradeChunk = uassertStatusOK(ChunkType::fromConfigBSON(chunk)); + bool historyIsAt40 = chunk[ChunkType::historyIsAt40()].booleanSafe(); + if (historyIsAt40) { + uassert( + ErrorCodes::Error(5760504), + str::stream() << "Chunk " << upgradeChunk.getName() << " in collection " << nss.ns() + << " indicates that it has been upgraded to version 4.0, but is " + "missing the history field. This indicates a corrupted routing " + "table and requires a manual intervention to be fixed.", + !upgradeChunk.getHistory().empty()); + continue; + } + + upgradeChunk.setVersion(newCollectionVersion); + newCollectionVersion.incMinor(); + changedShardIds.emplace(upgradeChunk.getShard()); + + // Construct the fresh history. + upgradeChunk.setHistory({ChunkHistory{validAfter, upgradeChunk.getShard()}}); + + // Set the 'historyIsAt40' field so that it gets skipped if the command is re-run + BSONObjBuilder chunkObjBuilder(upgradeChunk.toConfigBSON()); + chunkObjBuilder.appendBool(ChunkType::historyIsAt40(), true); + + // Run the update. + uassertStatusOK( + catalogClient->updateConfigDocument(opCtx, + ChunkType::ConfigNS, + BSON(ChunkType::name(upgradeChunk.getName())), + chunkObjBuilder.obj(), + false, + ShardingCatalogClient::kLocalWriteConcern)); + } + + // Wait for the writes to become majority committed so that the subsequent shard refreshes can + // see them + const auto clientOpTime = repl::ReplClientInfo::forClient(opCtx->getClient()).getLastOp(); + WriteConcernResult unusedWCResult; + uassertStatusOK(waitForWriteConcern( + opCtx, clientOpTime, ShardingCatalogClient::kMajorityWriteConcern, &unusedWCResult)); + + for (const auto& shardId : changedShardIds) { + auto shard = uassertStatusOK(shardRegistry->getShard(opCtx, shardId)); + uassertStatusOK( + Shard::CommandResponse::getEffectiveStatus(shard->runCommandWithFixedRetryAttempts( + opCtx, + ReadPreferenceSetting{ReadPreference::PrimaryOnly}, + "admin", + BSON("_flushRoutingTableCacheUpdates" << nss.ns()), + Shard::RetryPolicy::kIdempotent))); + } +} + StatusWith<ChunkVersion> ShardingCatalogManager::_findCollectionVersion( OperationContext* opCtx, const NamespaceString& nss, const OID& collectionEpoch) { auto const configShard = Grid::get(opCtx)->shardRegistry()->getConfigShard(); diff --git a/src/mongo/db/s/flush_routing_table_cache_updates_command.cpp b/src/mongo/db/s/flush_routing_table_cache_updates_command.cpp index 413f6cdf2ea..a5e44e89073 100644 --- a/src/mongo/db/s/flush_routing_table_cache_updates_command.cpp +++ b/src/mongo/db/s/flush_routing_table_cache_updates_command.cpp @@ -55,9 +55,9 @@ namespace { class FlushRoutingTableCacheUpdatesCmd final : public TypedCommand<FlushRoutingTableCacheUpdatesCmd> { public: - using Request = _flushRoutingTableCacheUpdates; + using Request = FlushRoutingTableCacheUpdates; - // Support deprecated name 'forceRoutingTableRefresh' for backwards compatibility with 3.6.0. + // Support deprecated name 'forceRoutingTableRefresh' for backwards compatibility with 4.0 FlushRoutingTableCacheUpdatesCmd() : TypedCommand<FlushRoutingTableCacheUpdatesCmd>(Request::kCommandName, "forceRoutingTableRefresh") {} @@ -104,11 +104,12 @@ public: uassertStatusOK(shardingState->canAcceptShardedCommands()); uassert(ErrorCodes::IllegalOperation, - "Can't issue _flushRoutingTableCacheUpdates from 'eval'", + str::stream() << "Can't issue " << Request::kCommandName << " from 'eval'", !opCtx->getClient()->isInDirectClient()); uassert(ErrorCodes::IllegalOperation, - "Can't call _flushRoutingTableCacheUpdates if in read-only mode", + str::stream() << "Can't call " << Request::kCommandName + << " if in read-only mode", !storageGlobalParams.readOnly); auto& oss = OperationShardingState::get(opCtx); diff --git a/src/mongo/s/catalog/type_chunk.cpp b/src/mongo/s/catalog/type_chunk.cpp index a70a2088f51..547d985c387 100644 --- a/src/mongo/s/catalog/type_chunk.cpp +++ b/src/mongo/s/catalog/type_chunk.cpp @@ -56,6 +56,7 @@ const BSONField<bool> ChunkType::jumbo("jumbo"); const BSONField<Date_t> ChunkType::lastmod("lastmod"); const BSONField<OID> ChunkType::epoch("lastmodEpoch"); const BSONField<BSONObj> ChunkType::history("history"); +const BSONField<bool> ChunkType::historyIsAt40("historyIsAt40"); namespace { diff --git a/src/mongo/s/catalog/type_chunk.h b/src/mongo/s/catalog/type_chunk.h index 72450bad707..9cb03eea92a 100644 --- a/src/mongo/s/catalog/type_chunk.h +++ b/src/mongo/s/catalog/type_chunk.h @@ -186,6 +186,7 @@ public: static const BSONField<Date_t> lastmod; static const BSONField<OID> epoch; static const BSONField<BSONObj> history; + static const BSONField<bool> historyIsAt40; ChunkType(); ChunkType(NamespaceString nss, ChunkRange range, ChunkVersion version, ShardId shardId); diff --git a/src/mongo/s/client/shard.h b/src/mongo/s/client/shard.h index 44a2c48c43b..faee4157968 100644 --- a/src/mongo/s/client/shard.h +++ b/src/mongo/s/client/shard.h @@ -40,11 +40,11 @@ #include "mongo/db/repl/read_concern_args.h" #include "mongo/executor/remote_command_response.h" #include "mongo/s/shard_id.h" +#include "mongo/s/write_ops/batched_command_request.h" +#include "mongo/s/write_ops/batched_command_response.h" namespace mongo { -class BatchedCommandRequest; -class BatchedCommandResponse; class OperationContext; class RemoteCommandTargeter; diff --git a/src/mongo/s/commands/SConscript b/src/mongo/s/commands/SConscript index f363155acee..8407084e202 100644 --- a/src/mongo/s/commands/SConscript +++ b/src/mongo/s/commands/SConscript @@ -72,6 +72,7 @@ env.Library( 'cluster_profile_cmd.cpp', 'cluster_remove_shard_cmd.cpp', 'cluster_remove_shard_from_zone_cmd.cpp', + 'cluster_repair_sharded_collection_chunks_history_cmd.cpp', 'cluster_repl_set_get_status_cmd.cpp', 'cluster_reset_error_cmd.cpp', 'cluster_restart_catalog_command.cpp', diff --git a/src/mongo/s/commands/cluster_repair_sharded_collection_chunks_history_cmd.cpp b/src/mongo/s/commands/cluster_repair_sharded_collection_chunks_history_cmd.cpp new file mode 100644 index 00000000000..22f9ccf7525 --- /dev/null +++ b/src/mongo/s/commands/cluster_repair_sharded_collection_chunks_history_cmd.cpp @@ -0,0 +1,122 @@ + +/** + * Copyright (C) 2021 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 + * <http://www.mongodb.com/licensing/server-side-public-license>. + * + * 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_LOG_DEFAULT_COMPONENT ::mongo::logger::LogComponent::kSharding + +#include "mongo/platform/basic.h" + +#include "mongo/db/audit.h" +#include "mongo/db/auth/action_set.h" +#include "mongo/db/auth/action_type.h" +#include "mongo/db/auth/authorization_manager.h" +#include "mongo/db/auth/authorization_session.h" +#include "mongo/db/client.h" +#include "mongo/db/commands.h" +#include "mongo/s/grid.h" + +namespace mongo { +namespace { + +class RepairShardedCollectionChunksHistoryCommand : public BasicCommand { +public: + RepairShardedCollectionChunksHistoryCommand() + : BasicCommand("repairShardedCollectionChunksHistory") {} + + AllowedOnSecondary secondaryAllowed(ServiceContext*) const override { + return AllowedOnSecondary::kAlways; + } + + bool adminOnly() const override { + return true; + } + + bool supportsWriteConcern(const BSONObj& cmd) const override { + return false; + } + + std::string help() const override { + return "Administrative command to repair the effects of SERVER-62065. If the collection " + "has been upgraded through a cluster comprised of binaries which do not contain " + "this command, the chunks cache collections on the shards will miss history " + "entries. This command will correct that and will mark such collections as " + "correctly repaired, so that a subsequent invocation will not cause any changes to " + "the routing information. In rare cases where the history entries are missing due " + "to corrupted restore, the 'force:true' parameter can be passed which will force " + "all history entries to be re-added."; + } + + // The command intentionally uses the permission control of split/mergeChunks since it only + // modifies the contents of chunk entries and increments the collection/shard versions without + // causing any data placement changes + Status checkAuthForCommand(Client* client, + const std::string& dbname, + const BSONObj& cmdObj) const override { + if (!AuthorizationSession::get(client)->isAuthorizedForActionsOnResource( + ResourcePattern::forExactNamespace(NamespaceString(parseNs(dbname, cmdObj))), + ActionType::splitChunk)) { + return Status(ErrorCodes::Unauthorized, "Unauthorized"); + } + return Status::OK(); + } + + std::string parseNs(const std::string& unusedDbName, const BSONObj& cmdObj) const override { + return CommandHelpers::parseNsFullyQualified(cmdObj); + } + + bool run(OperationContext* opCtx, + const std::string& unusedDbName, + const BSONObj& cmdObj, + BSONObjBuilder& result) override { + const NamespaceString nss{parseNs(unusedDbName, cmdObj)}; + + BSONObjBuilder cmdBuilder( + BSON("_configsvrRepairShardedCollectionChunksHistory" << nss.ns())); + if (cmdObj["force"].booleanSafe()) + cmdBuilder.appendBool("force", true); + + auto configShard = Grid::get(opCtx)->shardRegistry()->getConfigShard(); + auto cmdResponse = uassertStatusOK(configShard->runCommandWithFixedRetryAttempts( + opCtx, + ReadPreferenceSetting{ReadPreference::PrimaryOnly}, + "admin", + cmdBuilder.obj(), + Shard::RetryPolicy::kIdempotent)); + uassertStatusOK(cmdResponse.commandStatus); + + // Append any return value from the response, which the config server returned + CommandHelpers::filterCommandReplyForPassthrough(cmdResponse.response, &result); + + return true; + } + +} repairShardedCollectionChunksHistoryCommand; + +} // namespace +} // namespace mongo diff --git a/src/mongo/s/request_types/flush_routing_table_cache_updates.idl b/src/mongo/s/request_types/flush_routing_table_cache_updates.idl index 07981295bac..e5a27a58126 100644 --- a/src/mongo/s/request_types/flush_routing_table_cache_updates.idl +++ b/src/mongo/s/request_types/flush_routing_table_cache_updates.idl @@ -26,8 +26,6 @@ # it in the license file. # -# _flushRoutingTableCacheUpdates IDL File - global: cpp_namespace: "mongo" @@ -36,6 +34,7 @@ imports: commands: _flushRoutingTableCacheUpdates: + cpp_name: FlushRoutingTableCacheUpdates description: "An internal command to wait for the last routing table cache refresh for a particular namespace to be persisted to disk" strict: true namespace: type |