/** * Copyright (c) 2021 OceanBase * OceanBase CE is licensed under Mulan PubL v2. * You can use this software according to the terms and conditions of the Mulan PubL v2. * You may obtain a copy of Mulan PubL v2 at: * http://license.coscl.org.cn/MulanPubL-2.0 * THIS SOFTWARE IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND, * EITHER EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT, * MERCHANTABILITY OR FIT FOR A PARTICULAR PURPOSE. * See the Mulan PubL v2 for more details. */ #define USING_LOG_PREFIX SERVER #include "observer/ob_service.h" #include #include #include #include "share/ob_define.h" #include "lib/ob_running_mode.h" #include "lib/utility/utility.h" #include "lib/utility/ob_tracepoint.h" #include "lib/thread_local/ob_tsi_factory.h" #include "lib/utility/utility.h" #include "lib/time/ob_tsc_timestamp.h" #include "common/ob_partition_key.h" #include "common/ob_member_list.h" #include "common/ob_zone.h" #include "share/ob_version.h" #include "share/ob_version.h" #include "share/inner_table/ob_inner_table_schema.h" #include "share/partition_table/ob_partition_info.h" #include "share/partition_table/ob_partition_table_proxy.h" #include "share/partition_table/ob_partition_table_operator.h" #include "share/ob_tenant_mgr.h" #include "share/ob_sstable_checksum_operator.h" #include "share/ob_pg_partition_meta_table_operator.h" #include "share/ob_zone_table_operation.h" #include "storage/ob_partition_migrator.h" #include "storage/ob_partition_component_factory.h" #include "storage/ob_partition_migrator.h" #include "storage/ob_i_table.h" #include "storage/ob_pg_partition.h" #include "storage/ob_wrs_utils.h" // get_tenant_replica_wrs_info #include "storage/ob_partition_scheduler.h" #include "sql/optimizer/ob_opt_est_cost.h" #include "sql/optimizer/ob_join_order.h" #include "rootserver/ob_bootstrap.h" #include "observer/ob_server.h" #include "observer/ob_dump_task_generator.h" #include "observer/ob_server_schema_updater.h" #include "ob_server_event_history_table_operator.h" namespace oceanbase { using namespace common; using namespace rootserver; using namespace obrpc; using namespace share; using namespace share::schema; using namespace storage; namespace observer { static int64_t total_get_member_list_and_leader_cnt = 0; ObSchemaReleaseTimeTask::ObSchemaReleaseTimeTask() : schema_updater_(nullptr), is_inited_(false) {} int ObSchemaReleaseTimeTask::init(ObServerSchemaUpdater& schema_updater, int tg_id) { int ret = OB_SUCCESS; if (OB_UNLIKELY(is_inited_)) { ret = OB_INIT_TWICE; LOG_WARN("ObSchemaReleaseTimeTask has already been inited", K(ret)); } else { schema_updater_ = &schema_updater; is_inited_ = true; if (OB_FAIL(TG_SCHEDULE(tg_id, *this, REFRESH_INTERVAL, true /*schedule repeatly*/))) { LOG_WARN("fail to schedule task ObSchemaReleaseTimeTask", K(ret)); } } return ret; } void ObSchemaReleaseTimeTask::destroy() { is_inited_ = false; schema_updater_ = nullptr; } void ObSchemaReleaseTimeTask::runTimerTask() { int ret = OB_SUCCESS; if (OB_UNLIKELY(!is_inited_)) { ret = OB_NOT_INIT; LOG_WARN("ObSchemaReleaseTimeTask has not been inited", K(ret)); } else if (OB_ISNULL(schema_updater_)) { ret = OB_ERR_UNEXPECTED; LOG_WARN("ObSchemaReleaseTimeTask task got null ptr", K(ret)); } else if (OB_FAIL(schema_updater_->try_release_schema())) { LOG_WARN("ObSchemaReleaseTimeTask failed", K(ret)); } } ////////////////////////////////////// // here gctx may hasn't been initialized already ObService::ObService(const ObGlobalContext& gctx) : inited_(false), in_register_process_(false), service_started_(false), stopped_(false), schema_updater_(), partition_table_updater_(), index_status_report_queue_(), rebuild_flag_report_queue_(), pt_checker_(), lease_state_mgr_(), heartbeat_process_(gctx, schema_updater_, lease_state_mgr_), gctx_(gctx) {} ObService::~ObService() {} int ObService::init(common::ObMySQLProxy& sql_proxy) { int ret = OB_SUCCESS; const static int64_t INDEX_STATUS_REPORT_THREAD_CNT = 1; const static int64_t REBUILD_FLAG_REPORT_THREAD_CNT = 1; if (inited_) { ret = OB_INIT_TWICE; LOG_WARN("Oceanbase service has already init", K(ret)); } else if (!gctx_.is_inited()) { ret = OB_INVALID_ARGUMENT; LOG_WARN("gctx not init", "gctx inited", gctx_.is_inited(), K(ret)); } else if (OB_FAIL(heartbeat_process_.init())) { LOG_WARN("heartbeat_process_.init failed", K(ret)); } else if (OB_FAIL(schema_updater_.init(gctx_.self_addr_, gctx_.schema_service_))) { LOG_ERROR("client_manager_.initialize failed", "self_addr", gctx_.self_addr_, K(ret)); } else if (OB_FAIL(partition_table_updater_.init())) { LOG_WARN("init partition table updater failed", K(ret)); } else if (OB_FAIL(checksum_updater_.init())) { LOG_WARN("fail to init checksum updater", K(ret)); } else if (OB_FAIL(ObPGPartitionMTUpdater::get_instance().init())) { LOG_WARN("fail to init checksum updater", K(ret)); } else if (OB_FAIL(index_status_report_queue_.init(&index_updater_, INDEX_STATUS_REPORT_THREAD_CNT, !lib::is_mini_mode() ? ObPartitionTableUpdater::MAX_PARTITION_CNT : ObPartitionTableUpdater::MINI_MODE_MAX_PARTITION_CNT, "IdxStatRp"))) { LOG_WARN("init index status report queue failed", K(ret), LITERAL_K(INDEX_STATUS_REPORT_THREAD_CNT), "queue_size", static_cast(ObPartitionTableUpdater::MAX_PARTITION_CNT)); } else if (OB_FAIL(rebuild_flag_report_queue_.init(&rebuild_updater_, REBUILD_FLAG_REPORT_THREAD_CNT, !lib::is_mini_mode() ? ObPartitionTableUpdater::MAX_PARTITION_CNT : ObPartitionTableUpdater::MINI_MODE_MAX_PARTITION_CNT, "RebuFlagRp"))) { LOG_WARN("init rebuild flag report queue failed", K(ret), LITERAL_K(REBUILD_FLAG_REPORT_THREAD_CNT), "queue_size", static_cast(ObPartitionTableUpdater::MAX_PARTITION_CNT)); } else if (OB_FAIL(pt_checker_.init( *gctx_.pt_operator_, *gctx_.schema_service_, *gctx_.par_ser_, lib::TGDefIDs::ServerGTimer))) { LOG_WARN("init pt checker failed", K(ret)); } else if (OB_FAIL(SERVER_EVENT_INSTANCE.init(sql_proxy, gctx_.self_addr_))) { LOG_WARN("init server event history table failed", K(ret)); } else if (OB_FAIL(ObAllServerTracer::get_instance().init(lib::TGDefIDs::ServerGTimer, server_trace_task_))) { STORAGE_LOG(WARN, "fail to init ObAllServerTracer", K(ret)); } else if (OB_FAIL(OB_TSC_TIMESTAMP.init())) { LOG_WARN("fail to init tsc timestamp", K(ret)); } else if (OB_FAIL(schema_release_task_.init(schema_updater_, lib::TGDefIDs::ServerGTimer))) { LOG_WARN("fail to init schema release task", K(ret)); } else if (OB_FAIL(tenant_backup_task_updater_.init(sql_proxy))) { LOG_WARN("failed to init tenant backup task updater", K(ret)); } else if (OB_FAIL(pg_backup_task_updater_.init(sql_proxy))) { LOG_WARN("failed to init pg backup task updater", K(ret)); } else if (OB_FAIL(tenant_validate_task_updater_.init(sql_proxy))) { LOG_WARN("failed to init tenant validate task updater", K(ret)); } else if (OB_FAIL(pg_validate_task_updater_.init(sql_proxy))) { LOG_WARN("failed to init pg validate task updater", K(ret)); } else { inited_ = true; } return ret; } int ObService::register_self() { int ret = OB_SUCCESS; in_register_process_ = true; if (!inited_) { ret = OB_NOT_INIT; LOG_ERROR("service not initialized, can't register self", K(ret)); } else if (!lease_state_mgr_.is_inited()) { ret = OB_NOT_INIT; LOG_WARN("lease_state_mgr_ not init", K(ret)); } else if (OB_FAIL(lease_state_mgr_.register_self_busy_wait())) { LOG_WARN("register self failed", K(ret)); } else if (!lease_state_mgr_.is_valid_heartbeat()) { ret = OB_ERROR; LOG_ERROR("can't renew lease", K(ret), "heartbeat_expire_time", lease_state_mgr_.get_heartbeat_expire_time()); } else { in_register_process_ = false; service_started_ = true; SERVER_EVENT_ADD("observice", "register"); } return ret; } int ObService::start() { int ret = OB_SUCCESS; if (!inited_) { ret = OB_NOT_INIT; } // lease service if (OB_SUCC(ret)) { if (OB_FAIL(lease_state_mgr_.init(gctx_.rs_rpc_proxy_, gctx_.rs_mgr_, &heartbeat_process_, *this))) { LOG_ERROR("lease_state_mgr_.init failed", K(ret)); } else if (OB_FAIL(register_self())) { LOG_ERROR("register self failed", K(ret)); } else { LOG_INFO("[NOTICE] regist to rs success"); } } if (OB_SUCC(ret)) { if (OB_FAIL(report_replica())) { LOG_WARN("request report replica failed", K(ret)); } } if (OB_SUCC(ret)) { if (OB_FAIL(schedule_pt_check_task())) { LOG_WARN("schedule_pt_check_task failed", K(ret)); } } return ret; } void ObService::set_stop() { LOG_INFO("[NOTICE] observice need stop now"); lease_state_mgr_.set_stop(); } void ObService::stop() { if (!inited_) { LOG_WARN("not init"); } else { LOG_INFO("[NOTICE] start to stop"); SERVER_EVENT_ADD("observer", "stop"); service_started_ = false; stopped_ = true; schema_updater_.stop(); partition_table_updater_.stop(); checksum_updater_.stop(); ObPGPartitionMTUpdater::get_instance().stop(); index_status_report_queue_.stop(); rebuild_flag_report_queue_.stop(); pt_checker_.stop(); LOG_INFO("[NOTICE] observice finish stop"); } } void ObService::wait() { if (!inited_) { LOG_WARN("not init"); } else { schema_updater_.wait(); partition_table_updater_.wait(); checksum_updater_.wait(); ObPGPartitionMTUpdater::get_instance().wait(); index_status_report_queue_.wait(); rebuild_flag_report_queue_.wait(); } } int ObService::destroy() { int ret = OB_SUCCESS; if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else { schema_updater_.destroy(); lease_state_mgr_.destroy(); if (OB_FAIL(pt_checker_.destroy())) { LOG_WARN("pt_checker_ destroy failed", K(ret)); } SERVER_EVENT_INSTANCE.destroy(); LOG_INFO("server event table operator destroy"); } return ret; } int ObService::batch_get_role(const ObBatchGetRoleArg& arg, ObBatchGetRoleResult& result) { int ret = OB_SUCCESS; int64_t count = arg.keys_.count(); if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else if (OB_FAIL(result.results_.reserve(count))) { LOG_WARN("fail to init results", K(ret), K(arg)); } else { for (int64_t i = 0; OB_SUCC(ret) && i < count; i++) { const ObPartitionKey& key = arg.keys_.at(i); ObRole role; int res_ret = OB_SUCCESS; if (!key.is_valid()) { res_ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid argument", K(res_ret), K(key)); } else if (OB_SUCCESS != (res_ret = gctx_.par_ser_->get_role(key, role))) { LOG_WARN("fail to get role", K(res_ret), K(key)); } else if (!is_strong_leader(role)) { res_ret = OB_NOT_MASTER; } if (OB_FAIL(result.results_.push_back(res_ret))) { LOG_WARN("fail to push back ret", K(ret), K(res_ret)); } } } return ret; } int ObService::get_role(const ObPartitionKey& part_key, common::ObRole& role) { int ret = OB_SUCCESS; if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else if (!part_key.is_valid()) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid argument", K(part_key), K(ret)); } else if (gctx_.par_ser_->get_role(part_key, role)) { LOG_WARN("fail to get role", K(ret), K(part_key)); } return ret; } int ObService::get_leader_member(const ObPartitionKey& part_key, ObIArray& member_list) { int ret = OB_SUCCESS; if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else if (!part_key.is_valid()) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid argument", K(part_key), K(ret)); } else { ObMemberList ml; if (OB_FAIL(gctx_.par_ser_->get_leader_curr_member_list(part_key, ml))) { LOG_WARN("get leader current member list failed", K(ret)); } else if (ml.get_member_number() <= 0) { ret = OB_ERR_UNEXPECTED; LOG_WARN("invalid member list", K(ml), K(ret)); } else { member_list.reuse(); ObAddr addr; for (int64_t i = 0; OB_SUCC(ret) && i < ml.get_member_number(); ++i) { addr.reset(); if (OB_FAIL(ml.get_server_by_index(i, addr))) { LOG_WARN("get member failed", K(ret), K(i), K(ml)); } else if (!addr.is_valid()) { ret = OB_ERR_UNEXPECTED; LOG_WARN("invalid addr", K(addr), K(ret)); } else if (OB_FAIL(member_list.push_back(addr))) { LOG_WARN("push to array failed", K(ret)); } } } } return ret; } int ObService::update_baseline_schema_version(const int64_t schema_version) { int ret = OB_SUCCESS; ObMultiVersionSchemaService* schema_service = gctx_.schema_service_; if (schema_version <= 0) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid argument", KR(ret), K(schema_version)); } else if (OB_ISNULL(schema_service)) { ret = OB_ERR_UNEXPECTED; LOG_WARN("invalid schema service", KR(ret)); } else if (OB_FAIL(schema_service->update_baseline_schema_version(schema_version))) { LOG_WARN("fail to update baseline schema version", KR(ret), K(schema_version)); } else { LOG_INFO("update baseline schema version success", K(schema_version)); } return ret; } int ObService::get_member_list_and_leader(const common::ObPartitionKey& part_key, obrpc::ObMemberListAndLeaderArg& arg) { int ret = OB_SUCCESS; ObMemberList member_list; ObChildReplicaList children_list; ObAddr leader; ObRole role; ObAddr server; ObReplicaType replica_type = common::ObReplicaType::REPLICA_TYPE_MAX; ObReplicaProperty property; if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not inited", K(ret)); } else if (!part_key.is_valid()) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid args", K(ret), K(part_key)); } else if (OB_FAIL(gctx_.par_ser_->get_curr_leader_and_memberlist( part_key, leader, role, member_list, children_list, replica_type, property))) { LOG_WARN("failed to get get_curr_leader_and_memberlist", K(ret), K(part_key)); } else { arg.reset(); arg.leader_ = leader; arg.self_ = gctx_.self_addr_; arg.replica_type_ = replica_type; arg.property_ = property; arg.role_ = role; for (int64_t i = 0; OB_SUCC(ret) && i < member_list.get_member_number(); ++i) { if (OB_FAIL(member_list.get_server_by_index(i, server))) { LOG_WARN("failed to get server by index", K(ret), K(i)); } else if (OB_FAIL(arg.member_list_.push_back(server))) { LOG_WARN("failed to add member", K(ret)); } } } if (OB_SUCC(ret)) { if (OB_FAIL(arg.lower_list_.assign(children_list))) { LOG_WARN("fail to deep copy", K(ret)); } } static const int64_t STATISTICS_INTERVAL_US = 10000000; ATOMIC_INC(&total_get_member_list_and_leader_cnt); if (REACH_TIME_INTERVAL(STATISTICS_INTERVAL_US)) { LOG_INFO("get member_list_and_leader statistics", K(total_get_member_list_and_leader_cnt), "avg_rpc", total_get_member_list_and_leader_cnt / (STATISTICS_INTERVAL_US / 1000000)); ATOMIC_STORE(&total_get_member_list_and_leader_cnt, 0); } return ret; } int ObService::get_member_list_and_leader_v2( const common::ObPartitionKey& part_key, obrpc::ObGetMemberListAndLeaderResult& arg) { int ret = OB_SUCCESS; ObMemberList member_list; ObChildReplicaList children_list; ObAddr leader; ObRole role; ObMember member; ObReplicaType replica_type = common::ObReplicaType::REPLICA_TYPE_MAX; ObReplicaProperty property; if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not inited", K(ret)); } else if (!part_key.is_valid()) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid args", K(ret), K(part_key)); } else if (OB_FAIL(gctx_.par_ser_->get_curr_leader_and_memberlist( part_key, leader, role, member_list, children_list, replica_type, property))) { LOG_WARN("failed to get get_curr_leader_and_memberlist", K(ret), K(part_key)); } else { arg.reset(); arg.leader_ = leader; arg.self_ = gctx_.self_addr_; arg.replica_type_ = replica_type; arg.property_ = property; for (int64_t i = 0; OB_SUCC(ret) && i < member_list.get_member_number(); ++i) { if (OB_FAIL(member_list.get_member_by_index(i, member))) { LOG_WARN("failed to get server by index", K(ret), K(i)); } else if (OB_FAIL(arg.member_list_.push_back(member))) { LOG_WARN("failed to add member", K(ret)); } } } if (OB_SUCC(ret)) { if (OB_FAIL(arg.lower_list_.assign(children_list))) { LOG_WARN("fail to deep copy", K(ret)); } } static const int64_t STATISTICS_INTERVAL_US = 10000000; ATOMIC_INC(&total_get_member_list_and_leader_cnt); if (REACH_TIME_INTERVAL(STATISTICS_INTERVAL_US)) { LOG_INFO("get member_list_and_leader statistics", K(total_get_member_list_and_leader_cnt), "avg_rpc", total_get_member_list_and_leader_cnt / (STATISTICS_INTERVAL_US / 1000000)); ATOMIC_STORE(&total_get_member_list_and_leader_cnt, 0); } return ret; } int ObService::batch_get_member_list_and_leader( const obrpc::ObLocationRpcRenewArg& arg, obrpc::ObLocationRpcRenewResult& res) { int ret = OB_SUCCESS; if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not inited", K(ret)); } else if (!arg.is_valid()) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid args", K(ret), K(arg)); } else { ObMemberListAndLeaderArg result; for (int64_t i = 0; OB_SUCC(ret) && i < arg.keys_.count(); i++) { const ObPartitionKey& pkey = arg.keys_.at(i); result.reset(); int tmp_ret = OB_SUCCESS; if (OB_SUCCESS != (tmp_ret = get_member_list_and_leader(pkey, result))) { LOG_WARN("fail to get member_list and leader", K(tmp_ret), K(pkey)); } // always push back result if (OB_FAIL(res.results_.push_back(result))) { LOG_WARN("push back result failed", K(ret), K(pkey), K(result)); } } } return ret; } int ObService::get_pg_key(const ObPartitionKey& pkey, ObPGKey& pg_key) const { int ret = OB_SUCCESS; if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("service not inited", K(ret)); } else if (!pkey.is_valid()) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid argument", K(pkey), K(ret)); } else if (OB_ISNULL(gctx_.par_ser_)) { ret = OB_ERR_UNEXPECTED; LOG_WARN("partition_service is null", K(ret)); } else if (pkey.is_pg()) { pg_key = pkey; } else { ObPartitionGroupIndex& pg_index = gctx_.par_ser_->get_pg_index(); if (OB_FAIL(pg_index.get_pg_key(pkey, pg_key))) { if (OB_ENTRY_NOT_EXIST == ret) { pg_key = pkey; ret = OB_SUCCESS; } else { LOG_WARN("fail to get pg_key", K(ret), K(pkey)); } } } return ret; } int ObService::fill_partition_replica(const ObPGKey& pg_key, share::ObPartitionReplica& replica) { int ret = OB_SUCCESS; ObIPartitionGroup* part = NULL; ObIPartitionGroupGuard guard; if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("service not inited", K(ret)); } else if (!pg_key.is_valid()) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid argument", K(pg_key), K(ret)); } else if (OB_ISNULL(gctx_.par_ser_)) { ret = OB_ERR_UNEXPECTED; LOG_WARN("partition_service is null", K(ret)); } else { const ObPGMgr& pg_mgr = gctx_.par_ser_->get_pg_mgr(); if (OB_FAIL(pg_mgr.get_pg(pg_key, nullptr /*file_id*/, guard))) { LOG_WARN("fail to get pg guard", K(pg_key)); } else if (NULL == (part = guard.get_partition_group())) { ret = OB_ERR_UNEXPECTED; LOG_WARN("part is null", KP(part), K(ret)); } else if (OB_FAIL(fill_partition_replica(part, replica))) { LOG_WARN("failed to fill_partition_replica", K(ret), K(pg_key)); } } return ret; } int ObService::fill_partition_replica(storage::ObIPartitionGroup* part, share::ObPartitionReplica& replica) { int ret = OB_SUCCESS; if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("service not inited", K(ret)); } else if (OB_ISNULL(part)) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid argument", K(ret), KP(part)); } else { const ObPartitionKey& part_key = part->get_partition_key(); replica.reset(); replica.table_id_ = part_key.table_id_; replica.partition_id_ = part_key.get_partition_id(); replica.partition_cnt_ = part_key.get_partition_cnt(); replica.zone_ = gctx_.config_->zone.str(); replica.server_ = gctx_.self_addr_; replica.sql_port_ = gctx_.config_->mysql_port; int64_t leader_active_time = 0; if (!part->is_valid()) { ret = OB_INVALID_PARTITION; LOG_WARN("invalid partition", K(ret), K(part_key)); } else if (OB_ISNULL(part->get_log_service())) { ret = OB_ERR_UNEXPECTED; LOG_WARN("NULL partition log service", K(ret), K(part_key), KP(part)); } else if (OB_FAIL( part->get_log_service()->get_role_and_last_leader_active_time(replica.role_, leader_active_time))) { LOG_WARN("get partition role failed", K(ret), K(part_key)); // After the version number is greater than or equal to 2.1, // the observer schedules the rebuild task by itself, // but still needs to report the rebuild tag, // because it needs to skip the replica in rebuild while major merge } else if (OB_FAIL(part->get_log_service()->is_need_rebuild(replica.rebuild_))) { LOG_WARN("get partition rebuild state failed", K(ret), K(part_key)); } else if (OB_FAIL(part->get_log_service()->get_replica_num(replica.quorum_))) { LOG_WARN("get partition quorum failed", K(ret), K(part_key)); } else { if (replica.is_leader_like()) { replica.to_leader_time_ = leader_active_time; } if (OB_FAIL(part->fill_replica(replica))) { LOG_WARN("fill replica failed", K(ret), K(part_key)); // temporary online replica doesn't have sstable if (OB_ENTRY_NOT_EXIST == ret) { ret = OB_SUCCESS; } } if (OB_FAIL(ret)) { } else { ObMemberList member_list; if (OB_FAIL(part->get_curr_member_list_for_report(member_list))) { LOG_WARN("get_curr_member_list_for_report failed", K(ret)); } else { ObMember member; for (int64_t i = 0; OB_SUCC(ret) && i < member_list.get_member_number(); ++i) { member.reset(); if (OB_FAIL(member_list.get_member_by_index(i, member))) { LOG_WARN("get_member_by_index failed", K(i), K(ret)); } else if (!member.is_valid()) { ret = OB_ERR_UNEXPECTED; LOG_WARN("member is invalid", K(member), K(ret)); } else { if (OB_FAIL(replica.member_list_.push_back( ObPartitionReplica::Member(member.get_server(), member.get_timestamp())))) { LOG_WARN("push_back failed", K(ret)); } } } } } } if (REACH_TIME_INTERVAL(1000 * 1000)) { ObTaskController::get().allow_next_syslog(); LOG_INFO("finish fill partition replica", K(ret), K(part_key), K(replica)); } else { ObTaskController::get().allow_next_syslog(); LOG_TRACE("finish fill partition replica", K(ret), K(part_key), K(replica)); } } return ret; } const ObAddr& ObService::get_self_addr() { return gctx_.self_addr_; } int ObService::fill_partition_table_update_task(const common::ObPartitionKey& pkey, ObPGPartitionMTUpdateItem& item) { int ret = OB_SUCCESS; if (OB_UNLIKELY(!inited_)) { ret = OB_NOT_INIT; LOG_WARN("ObService has not been inited", K(ret)); } else if (OB_UNLIKELY(!pkey.is_valid()) || pkey.is_pg()) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid arguments", K(ret), K(pkey)); } else { ObIPartitionGroupGuard guard; ObReportStatus report_status; common::ObRole role; ObReplicaStatus status; int64_t leader_active_time = 0; int checksum = 0; if (OB_FAIL(gctx_.par_ser_->get_partition(pkey, guard))) { LOG_WARN("fail to get partition", K(ret), K(pkey)); } else if (OB_ISNULL(guard.get_partition_group())) { ret = OB_ERR_UNEXPECTED; LOG_WARN("error unexpected, partition must not be NULL", K(ret)); } else if (OB_FAIL(guard.get_partition_group()->fill_pg_partition_replica(pkey, status, report_status))) { LOG_WARN("get pg partition fail", K(ret), K(pkey)); } else if (OB_FAIL(guard.get_partition_group()->get_log_service()->get_role_and_last_leader_active_time( role, leader_active_time))) { LOG_WARN("get partition role failed", K(ret), K(pkey)); } else { item.tenant_id_ = pkey.get_tenant_id(); item.table_id_ = pkey.get_table_id(); item.partition_id_ = pkey.get_partition_id(); item.svr_ip_ = GCTX.self_addr_; item.role_ = role; item.status_ = status; item.replica_type_ = guard.get_partition_group()->get_replica_type(); item.row_count_ = report_status.row_count_; item.data_size_ = report_status.data_size_; item.required_size_ = report_status.required_size_; item.data_version_ = report_status.data_version_; item.data_checksum_ = report_status.data_checksum_; } } if (OB_SUCC(ret)) { LOG_INFO("fill pg partition replica success", K(pkey), K(item)); } return ret; } int ObService::fill_checksum(const common::ObPartitionKey& pkey, const uint64_t sstable_id, const int sstable_type, const ObSSTableChecksumUpdateType update_type, common::ObIArray& data_checksum_items, common::ObIArray& column_checksum_items) { int ret = OB_SUCCESS; if (OB_UNLIKELY(!inited_)) { ret = OB_NOT_INIT; LOG_WARN("ObService has not been inited", K(ret)); } else if (OB_UNLIKELY(!pkey.is_valid())) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid arguments", K(ret), K(pkey)); } else { ObPartitionStorage* storage = NULL; ObIPartitionGroupGuard guard; ObPGPartitionGuard pg_partition_guard; ObSEArray sstable_ids; ObSchemaGetterGuard schema_guard; const ObTableSchema* main_table_schema = NULL; ObSEArray simple_index_infos; const uint64_t fetch_tenant_id = is_inner_table(pkey.get_table_id()) ? OB_SYS_TENANT_ID : pkey.get_tenant_id(); if (OB_FAIL(gctx_.par_ser_->get_partition(pkey, guard))) { LOG_WARN("fail to get partition", K(ret), K(pkey)); } else if (OB_ISNULL(guard.get_partition_group())) { ret = OB_ERR_UNEXPECTED; LOG_WARN("error unexpected, partition must not be NULL", K(ret)); } else if (OB_FAIL(guard.get_partition_group()->get_pg_partition(pkey, pg_partition_guard))) { LOG_WARN("get pg partition fail", K(ret), K(pkey)); } else if (OB_ISNULL(pg_partition_guard.get_pg_partition())) { ret = OB_ENTRY_NOT_EXIST; LOG_WARN("pg partition not exist", K(ret), K(pkey)); } else if (OB_ISNULL( storage = static_cast(pg_partition_guard.get_pg_partition()->get_storage()))) { ret = OB_ERR_UNEXPECTED; LOG_WARN("error unexpected, partition storage must not be NULL", K(ret)); } else if (OB_FAIL(ObMultiVersionSchemaService::get_instance().get_tenant_full_schema_guard( fetch_tenant_id, schema_guard))) { STORAGE_LOG(WARN, "fail to get schema guard", K(ret), K(fetch_tenant_id)); } else if (OB_FAIL(schema_guard.get_table_schema(pkey.get_table_id(), main_table_schema))) { STORAGE_LOG(WARN, "fail to get table schema", K(ret), K(pkey)); } else { bool need_report_column_checksum = false; if (nullptr == main_table_schema) { need_report_column_checksum = true; } else { if (main_table_schema->is_global_index_table()) { need_report_column_checksum = true; } else if (OB_FAIL(main_table_schema->get_simple_index_infos(simple_index_infos))) { STORAGE_LOG(WARN, "fail to get index tid array", K(ret)); } else { const ObSimpleTableSchemaV2* index_schema = nullptr; for (int64_t i = 0; OB_SUCC(ret) && i < simple_index_infos.count() && !need_report_column_checksum; ++i) { if (OB_FAIL(schema_guard.get_table_schema(simple_index_infos.at(i).table_id_, index_schema))) { STORAGE_LOG(WARN, "fail to get table schema", K(ret)); } else if (OB_ISNULL(index_schema)) { ret = OB_TABLE_NOT_EXIST; STORAGE_LOG(WARN, "fail to get index schema", K(ret), K(simple_index_infos.at(i).table_id_)); } else if (index_schema->is_global_index_table()) { need_report_column_checksum = true; } } } } if (OB_SUCC(ret)) { const bool need_update_data_checksum = observer::need_update_data_checksum(update_type); const bool need_update_column_checksum = need_report_column_checksum && observer::need_update_column_checksum(update_type); if (OB_INVALID_ID == sstable_id) { if (OB_FAIL(storage->get_partition_store().get_all_table_ids(sstable_ids))) { LOG_WARN("fail to get all table ids", K(ret)); } } else { if (OB_FAIL(sstable_ids.push_back(sstable_id))) { LOG_WARN("fail to push back sstable id", K(ret)); } } for (int64_t i = 0; OB_SUCC(ret) && i < sstable_ids.count(); ++i) { ObSSTableChecksumItem item; ObSSTableDataChecksumItem& data_checksum = item.data_checksum_; data_checksum.tenant_id_ = extract_tenant_id(pkey.get_table_id()); data_checksum.data_table_id_ = pkey.get_table_id(); data_checksum.partition_id_ = pkey.get_partition_id(); data_checksum.server_ = GCTX.self_addr_; if (OB_INVALID_ID == sstable_id) { const int table_types[] = { ObITable::MAJOR_SSTABLE, ObITable::MINOR_SSTABLE, ObITable::MULTI_VERSION_MINOR_SSTABLE}; for (int64_t j = 0; OB_SUCC(ret) && j < sizeof(table_types) / sizeof(int); ++j) { item.reset(); if (OB_FAIL(storage->fill_checksum(sstable_ids.at(i), table_types[j], item))) { if (OB_ENTRY_NOT_EXIST != ret) { LOG_WARN("fail to fill checksum", K(ret), "sstable_id", sstable_ids.at(i)); } else { ret = OB_SUCCESS; } } else if (need_update_data_checksum && OB_FAIL(data_checksum_items.push_back(item.data_checksum_))) { LOG_WARN("fail to push back item", K(ret)); } else if (need_update_column_checksum) { for (int64_t k = 0; OB_SUCC(ret) && k < item.column_checksum_.count(); ++k) { if (OB_FAIL(column_checksum_items.push_back(item.column_checksum_.at(k)))) { LOG_WARN("fail to push back column checksum item", K(ret)); } } } } } else { if (OB_FAIL(storage->fill_checksum(sstable_ids.at(i), sstable_type, item))) { if (OB_ENTRY_NOT_EXIST != ret) { LOG_WARN("fail to fill checksum", K(ret), "sstable_id", sstable_ids.at(i)); } else { ret = OB_SUCCESS; } } else if (need_update_data_checksum && OB_FAIL(data_checksum_items.push_back(item.data_checksum_))) { LOG_WARN("fail to push back item", K(ret)); } else if (need_update_column_checksum) { for (int64_t k = 0; OB_SUCC(ret) && k < item.column_checksum_.count(); ++k) { if (OB_FAIL(column_checksum_items.push_back(item.column_checksum_.at(k)))) { LOG_WARN("fail to push back column checksum item", K(ret)); } } } } } } } } return ret; } int ObService::pt_sync_update(const ObPartitionKey& part_key) { int ret = OB_SUCCESS; if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else if (!part_key.is_valid()) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid argument", K(part_key), K(ret)); } else if (OB_FAIL(partition_table_updater_.sync_update(part_key))) { LOG_WARN("async_update failed", K(part_key), K(ret)); } return ret; } void ObService::submit_pg_pt_update_task(const ObPartitionArray& pg_partition_keys) { int ret = OB_SUCCESS; const int64_t version = 0; if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else { // Batch update all partition information under PG for (int64_t i = 0; OB_SUCC(ret) && i < pg_partition_keys.count(); ++i) { if (OB_FAIL(ObPGPartitionMTUpdater::get_instance().add_task( pg_partition_keys.at(i), ObPGPartitionMTUpdateType::INSERT_ON_UPDATE, version, GCTX.self_addr_))) { LOG_WARN("fail to async update pg partition meta table", K(ret), K(pg_partition_keys.at(i))); } } if (OB_SUCC(ret)) { LOG_INFO("async update pg partition meta table success", K(pg_partition_keys)); } } UNUSED(ret); } int ObService::submit_pt_update_role_task(const ObPartitionKey& pkey) { int ret = OB_SUCCESS; if (OB_UNLIKELY(!inited_)) { ret = OB_NOT_INIT; LOG_WARN("not init", KR(ret)); } else if (OB_UNLIKELY(!pkey.is_valid())) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid argument", KR(ret), K(pkey)); } else if (OB_FAIL(partition_table_updater_.async_update_partition_role(pkey))) { LOG_WARN("fail to async update partition role", KR(ret), K(pkey)); } else { // do nothing } return ret; } int ObService::submit_pt_update_task( const ObPartitionKey& part_key, const bool need_report_checksum, const bool with_role) { int ret = OB_SUCCESS; const bool is_remove = false; const int64_t version = 0; const ObSSTableChecksumUpdateType update_type = ObSSTableChecksumUpdateType::UPDATE_ALL; if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else if (!part_key.is_valid()) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid argument", K(part_key), K(ret)); } else if (OB_FAIL(partition_table_updater_.async_update(part_key, with_role))) { LOG_WARN("async_update failed", K(part_key), K(ret)); } else if (need_report_checksum && !part_key.is_pg() && OB_FAIL(checksum_updater_.add_task(part_key, is_remove, update_type))) { LOG_WARN("fail to async update sstable checksum", K(ret)); } else { // do nothing } return ret; } int ObService::submit_checksum_update_task(const ObPartitionKey& part_key, const uint64_t sstable_id, const int sstable_type, const ObSSTableChecksumUpdateType update_type, const bool task_need_batch) { int ret = OB_SUCCESS; if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else if (!part_key.is_valid() || !is_valid_checksum_update_type(update_type)) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid argument", K(part_key), K(ret), K(part_key), K(sstable_id), K(sstable_type), K(update_type)); } else if (OB_FAIL(checksum_updater_.add_task( part_key, false /*is remove*/, update_type, sstable_id, sstable_type, task_need_batch))) { LOG_WARN("fail to async update sstable checksum", K(ret)); } else if (!part_key.is_pg()) { const int64_t version = 0; if (OB_FAIL(ObPGPartitionMTUpdater::get_instance().add_task( part_key, ObPGPartitionMTUpdateType::INSERT_ON_UPDATE, version, GCTX.self_addr_))) { LOG_WARN("fail to async update pg partition meta table", K(ret), K(part_key)); } } return ret; } int ObService::submit_async_refresh_schema_task(const uint64_t tenant_id, const int64_t schema_version) { int ret = OB_SUCCESS; if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else if (OB_INVALID_TENANT_ID == tenant_id || OB_INVALID_ID == tenant_id || !ObSchemaService::is_formal_version(schema_version)) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid arg", K(ret), K(tenant_id), K(schema_version)); } else if (OB_FAIL(schema_updater_.async_refresh_schema(tenant_id, schema_version))) { LOG_WARN("fail to async refresh schema", K(ret), K(tenant_id), K(schema_version)); } return ret; } int ObService::report_rebuild_replica(const common::ObPartitionKey& part_key, const common::ObAddr& server, const storage::ObRebuildSwitch& rebuild_switch) { int ret = OB_SUCCESS; bool rebuild = true; if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not inited", K(ret)); } else if (!part_key.is_valid() || !server.is_valid()) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid argument", K(ret), K(part_key), K(server)); } else if (OB_ISNULL(gctx_.pt_operator_)) { ret = OB_ERR_UNEXPECTED; LOG_WARN("NULL partition table operator", K(ret)); } else { switch (rebuild_switch) { case storage::OB_REBUILD_ON: rebuild = true; break; case storage::OB_REBUILD_OFF: rebuild = false; break; default: ret = OB_INVALID_ARGUMENT; LOG_WARN("unkown rebuild switch", K(part_key), K(server), K(rebuild_switch), K(ret)); } if (OB_SUCC(ret)) { if (OB_FAIL(gctx_.pt_operator_->update_rebuild_flag( part_key.get_table_id(), part_key.get_partition_id(), server, rebuild))) { LOG_WARN("update partition table rebuild flag failed", K(ret), K(part_key), K(server), K(rebuild)); } } } return ret; } int ObService::report_rebuild_replica_async(const common::ObPartitionKey& part_key, const common::ObAddr& server, const storage::ObRebuildSwitch& rebuild_switch) { int ret = OB_SUCCESS; if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not inited", K(ret)); } else if (!part_key.is_valid() || !server.is_valid() || OB_REBUILD_INVALID == rebuild_switch) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid argument", K(ret), K(part_key), K(server), K(rebuild_switch)); } else { ObRebuildFlagReporter task; if (OB_FAIL(task.init(part_key, server, rebuild_switch))) { LOG_WARN("init rebulid flag report task failed", K(ret), K(part_key), K(server), K(rebuild_switch)); } else { ret = rebuild_flag_report_queue_.add(task); LOG_INFO("submit rebuild flag report task", K(ret), K(task)); if (OB_FAIL(ret)) { if (OB_EAGAIN == ret) { ret = OB_SUCCESS; } else { LOG_WARN("submit rebuild flag report task failed", K(ret), K(task)); } } } } return ret; } int ObService::schedule_pt_check_task() { int ret = OB_SUCCESS; if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else { // to make different start check partition table task in different time const int64_t interval_in_sec = GCONF.partition_table_check_interval / 1000000; const int64_t delay = (random() % interval_in_sec) * 1000000; const int64_t repeat = false; if (OB_FAIL(TG_SCHEDULE(lib::TGDefIDs::ServerGTimer, pt_checker_, delay, repeat))) { LOG_WARN("schedule partition table check task failed", K(delay), K(repeat), K(ret)); } } return ret; } int ObService::report_merge_finished(int64_t frozen_version) { int ret = OB_SUCCESS; if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else if (frozen_version <= 0) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid argument", K(ret), K(frozen_version)); } else { if (frozen_version > 0) { if (OB_FAIL(partition_table_updater_.sync_merge_finish(frozen_version))) { LOG_WARN("add report merge finish task failed", K(ret), K(frozen_version)); } SERVER_EVENT_ADD("daily_merge", "report_merge_finished", K(ret), K(frozen_version)); } } return ret; } int ObService::report_merge_error(const common::ObPartitionKey& part_key, const int error_code) { int ret = OB_SUCCESS; LOG_WARN("merge error", K(part_key), K(error_code)); if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else if (!part_key.is_valid() || OB_SUCCESS == error_code) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid argument", K(ret), K(part_key), K(error_code)); } else { ObMergeErrorArg arg; arg.partition_key_ = part_key; arg.server_ = gctx_.self_addr_; arg.error_code_ = error_code; obrpc::ObCommonRpcProxy rs_rpc = *gctx_.rs_rpc_proxy_; rs_rpc.set_rs_mgr(*gctx_.rs_mgr_); if (OB_FAIL(rs_rpc.merge_error(arg))) { LOG_WARN("report merge error failed", K(ret), K(arg)); } SERVER_EVENT_ADD("daily_merge", "report_merge_error", K(ret), K(error_code), "partition", part_key); } return ret; } int ObService::report_local_index_build_complete(const common::ObPartitionKey& part_key, const uint64_t index_id, const share::schema::ObIndexStatus index_status, const int32_t ret_code) { int ret = OB_SUCCESS; if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else if (!part_key.is_valid() || OB_INVALID_ID == index_id || index_status <= INDEX_STATUS_NOT_FOUND || index_status >= INDEX_STATUS_MAX) { // don't check ret_code here ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid argument", K(part_key), K(index_id), K(index_status), K(ret)); } else { ObIndexStatusReporter task; if (OB_FAIL(task.init(part_key, get_self_addr(), index_id, index_status, ret_code, *gctx_.sql_proxy_))) { LOG_WARN("init index status report task failed", K(ret), K(part_key), K(index_id), K(ret_code)); } else { ret = index_status_report_queue_.add(task); LOG_INFO("submit index status report task", K(ret), K(task)); if (OB_FAIL(ret)) { if (OB_EAGAIN == ret) { ret = OB_SUCCESS; } else { LOG_WARN("submit index status report task failed", K(ret), K(task)); } } } SERVER_EVENT_ADD("daily_merge", "index_build_complete", K(ret)); if (OB_SUCCESS != ret_code) { LOG_ERROR("Fail to build local index, ", K(part_key), K(index_id), K(index_status), K(ret_code)); } } return ret; } int ObService::submit_pt_remove_task(const ObPartitionKey& part_key) { int ret = OB_SUCCESS; const int64_t version = 0; if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else if (!part_key.is_valid()) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid part_key", K(part_key), K(ret)); } else if (OB_FAIL(partition_table_updater_.async_remove(part_key))) { LOG_WARN("async_remove failed", K(part_key), K(ret)); } else if (OB_FAIL(checksum_updater_.add_task(part_key, true /*is remove*/))) { LOG_WARN("fail to add task", K(ret), K(part_key)); } else if (part_key.is_pg()) { // do nothing } else if (OB_FAIL(ObPGPartitionMTUpdater::get_instance().add_task( part_key, ObPGPartitionMTUpdateType::DELETE, version, GCTX.self_addr_))) { LOG_WARN("fail to async delete pg partition meta table", K(ret), K(part_key)); } else { LOG_INFO("async delete pg partition meta table success", K(part_key)); } return ret; } int ObService::reach_partition_limit(const obrpc::ObReachPartitionLimitArg& arg) { int ret = OB_SUCCESS; if (OB_UNLIKELY(!inited_)) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else if (OB_UNLIKELY(!arg.is_valid())) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid argument", K(ret), K(arg)); } else if (OB_ISNULL(gctx_.par_ser_)) { ret = OB_ERR_UNEXPECTED; LOG_WARN("partition service ptr is null", K(ret)); } else if (gctx_.par_ser_->reach_tenant_partition_limit(arg.batch_cnt_, arg.tenant_id_, arg.is_pg_arg_)) { ret = OB_TOO_MANY_PARTITIONS_ERROR; LOG_WARN("too many partitions on this server", K(ret), K(arg)); } return ret; } // should return success if all partition have merge to specific frozen_version int ObService::check_frozen_version(const obrpc::ObCheckFrozenVersionArg& arg) { LOG_INFO("receive check frozen version request", K(arg)); int ret = OB_SUCCESS; ObPartitionScheduler& scheduler = ObPartitionScheduler::get_instance(); int64_t last_merged_version = scheduler.get_merged_version(); if (OB_UNLIKELY(!inited_)) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else if (OB_UNLIKELY(!arg.is_valid())) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid argument", K(ret), K(arg)); } else if (arg.frozen_version_ != last_merged_version) { ret = OB_ERR_CHECK_DROP_COLUMN_FAILED; LOG_WARN("last merged version not match", K(ret), K(arg), K(last_merged_version)); } return ret; } int ObService::get_min_sstable_schema_version( const obrpc::ObGetMinSSTableSchemaVersionArg& arg, obrpc::ObGetMinSSTableSchemaVersionRes& result) { int ret = OB_SUCCESS; ObMultiVersionSchemaService* schema_service = gctx_.schema_service_; if (OB_UNLIKELY(!inited_)) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else if (OB_UNLIKELY(!arg.is_valid())) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid argument", K(ret), K(arg)); } else if (OB_ISNULL(schema_service)) { ret = OB_ERR_UNEXPECTED; LOG_WARN("schema service is null", K(ret)); } else { for (int i = 0; OB_SUCC(ret) && i < arg.tenant_id_arg_list_.size(); ++i) { // The minimum schema_version used by storage will increase with the major version, // storage only need to keep schema history used by a certain number major version. // min_schema_version = scheduler.get_min_schema_version(arg.tenant_id_arg_list_.at(i)); int tmp_ret = OB_SUCCESS; const uint64_t tenant_id = arg.tenant_id_arg_list_.at(i); int64_t min_schema_version = 0; if (OB_SUCCESS != (tmp_ret = schema_service->get_recycle_schema_version(tenant_id, min_schema_version))) { min_schema_version = OB_INVALID_VERSION; LOG_WARN("fail to get recycle schema version", K(tmp_ret), K(tenant_id)); } else { ObPartitionScheduler& scheduler = ObPartitionScheduler::get_instance(); min_schema_version = min(min_schema_version, scheduler.get_merged_schema_version(tenant_id)); } if (OB_FAIL(result.ret_list_.push_back(min_schema_version))) { LOG_WARN("push error", K(ret), K(arg)); } } } return ret; } int ObService::create_partition(const obrpc::ObCreatePartitionArg& arg) { UNUSED(arg); return OB_NOT_SUPPORTED; } int ObService::check_partition_need_update_pt_( const obrpc::ObCreatePartitionBatchArg& batch_arg, obrpc::ObCreatePartitionBatchRes& batch_res, bool& need_update) { int ret = OB_SUCCESS; need_update = true; if (batch_arg.args_.count() <= 0) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid argument", K(batch_arg)); } else if (batch_arg.args_.count() != batch_res.ret_list_.count()) { ret = OB_ERR_UNEXPECTED; LOG_WARN("result count mismatch with argument count", K(ret), "res_count", batch_res.ret_list_.count(), "arg_count", batch_arg.args_.count()); } else { // for record whether the first partition needs to update meta table bool first_need_sstable = false; const ObCreatePartitionArg& first = batch_arg.args_.at(0); if (OB_FAIL(first.check_need_create_sstable(first_need_sstable))) { LOG_WARN("fail to check need create sstable", K(ret)); } else if (is_sys_table(first.partition_key_.get_table_id())) { // sys_table updated to all_core_table need_update = false; } else { // do nothing } // Check whether the other partitions are the same as the first partition FOREACH_X(arg, batch_arg.args_, OB_SUCC(ret) && need_update) { bool this_need_sstable = false; if (OB_FAIL(arg->check_need_create_sstable(this_need_sstable))) { LOG_WARN("fail to check need create sstable", K(ret)); } else if (arg->partition_key_.get_table_id() != first.partition_key_.get_table_id() || arg->memstore_version_ != first.memstore_version_ || arg->replica_type_ != first.replica_type_ || first_need_sstable != this_need_sstable) { need_update = false; } else { // do nothing } } } return ret; } int ObService::create_partition_batch( const obrpc::ObCreatePartitionBatchArg& batch_arg, obrpc::ObCreatePartitionBatchRes& batch_res) { int ret = OB_SUCCESS; DEBUG_SYNC(DEFORE_OBS_CREATE_PARTITION); int64_t start_timestamp = ObTimeUtility::current_time(); LOG_INFO("receive batch create partitions request", K(start_timestamp), K(batch_arg.args_.count()), K(batch_arg)); int64_t partition_count = 0; if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else if (!batch_arg.is_valid()) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid batch arg", K(batch_arg), K(ret)); } else if (OB_ISNULL(gctx_.par_ser_) || OB_ISNULL(gctx_.sql_proxy_)) { ret = OB_ERR_UNEXPECTED; LOG_WARN("NULL partition service or sql proxy", K(ret), "partition_service", gctx_.par_ser_, "sql_proxy", gctx_.sql_proxy_); } else if (OB_FAIL(gctx_.par_ser_->get_partition_count(partition_count))) { LOG_WARN("fail to get partition count", K(ret)); } else { const obrpc::ObCreatePartitionArg& arg = batch_arg.args_.at(0); // add partition to pg if (arg.is_create_pg_partition()) { if (OB_FAIL(gctx_.par_ser_->create_batch_pg_partitions(batch_arg.args_, batch_res.ret_list_))) { LOG_WARN("failed to create batch partition groups.", K(ret)); } // create partition group / stand alone partition } else if (OB_FAIL(gctx_.par_ser_->create_batch_partition_groups(batch_arg.args_, batch_res.ret_list_))) { LOG_WARN("failed to create batch partition groups.", K(ret)); } else { // do nothing } } int64_t end_time = ObTimeUtility::current_time(); int64_t cost_time = end_time - start_timestamp; LOG_INFO("create partition register end", K(ret), K(cost_time)); // update partition table if (OB_SUCC(ret)) { bool need_update = true; if (OB_FAIL(check_partition_need_update_pt_(batch_arg, batch_res, need_update))) { LOG_WARN("check partition need update partition table error", K(ret), K(batch_arg), K(batch_res)); } else if (need_update) { bool first_need_sstable = false; const ObCreatePartitionArg& first = batch_arg.args_.at(0); if (OB_FAIL(first.check_need_create_sstable(first_need_sstable))) { LOG_WARN("fail to check need create sstable", K(ret)); } else { // update __all_tenant_meta_table int tmp_ret = OB_SUCCESS; int64_t data_version = first_need_sstable ? first.memstore_version_ - 1 : 0; // Those situations need to update sstable_checksum: // 1. standalone partition // 2. normal partition under pg; if (OB_SUCCESS != (tmp_ret = sync_report_replica_info(data_version, first, batch_arg))) { LOG_INFO("partition table sync report fail or timeout, need start async report", KR(tmp_ret), K(batch_arg), K(batch_res)); // It is failed to report meta_table, start an asynchronous task to update in the background FOREACH_X(arg, batch_arg.args_, OB_SUCC(ret)) { ObPartitionArray pkeys; if (OB_FAIL(submit_pt_update_task(arg->partition_key_))) { LOG_WARN("submit partition table async update task failed", K(ret), K(arg->partition_key_), "data_version", arg->memstore_version_ - 1); } else if (arg->partition_key_.is_pg()) { // do nothing } else if (OB_FAIL(pkeys.push_back(arg->partition_key_))) { LOG_WARN("pkeys push back error", K(ret), K(arg->partition_key_)); } else { // do nothing } if (OB_SUCC(ret) && pkeys.count() > 0) { // update partition meta table without consider the error codes submit_pg_pt_update_task(pkeys); } } } else { // It is success to update meta table, need to update all_sstable_checksum asynchronously const bool is_remove = false; const int64_t version = 0; const ObSSTableChecksumUpdateType update_type = ObSSTableChecksumUpdateType::UPDATE_ALL; FOREACH_X(arg, batch_arg.args_, OB_SUCC(ret)) { if (!arg->partition_key_.is_pg()) { if (OB_FAIL(checksum_updater_.add_task(arg->partition_key_, is_remove, update_type))) { LOG_WARN("fail to async update sstable checksum", K(ret), K(arg->partition_key_)); } else if (OB_FAIL(ObPGPartitionMTUpdater::get_instance().add_task(arg->partition_key_, ObPGPartitionMTUpdateType::INSERT_ON_UPDATE, version, GCTX.self_addr_))) { LOG_WARN("fail to async update sstable checksum", K(ret), K(arg->partition_key_)); } else { // do nothing } } } } } } else { FOREACH_X(arg, batch_arg.args_, OB_SUCC(ret)) { // RS not in service (in bootstrap) while creating __all_core_table partition, // update replica not need (always fail). if (combine_id(OB_SYS_TENANT_ID, OB_ALL_CORE_TABLE_TID) != arg->partition_key_.get_table_id()) { ObPartitionArray pkeys; if (OB_FAIL(submit_pt_update_task(arg->partition_key_))) { LOG_WARN("submit partition table async update task failed", K(ret), "partition_key", arg->partition_key_, "data_version", arg->memstore_version_ - 1); } else if (arg->partition_key_.is_pg()) { // do nothing } else if (OB_FAIL(pkeys.push_back(arg->partition_key_))) { LOG_WARN("pkeys push back error", K(ret), K(arg->partition_key_)); } else { // update partition meta table without consider the error code submit_pg_pt_update_task(pkeys); } } } } } int64_t cost = ObTimeUtility::current_time() - start_timestamp; int64_t update_pt_cost = ObTimeUtility::current_time() - end_time; LOG_INFO("create partition batch finish", K(ret), "partition count", batch_arg.args_.count(), K(cost), K(update_pt_cost), K(batch_res.ret_list_.count()), K(batch_res)); if (batch_arg.args_.count() == 1) { SERVER_EVENT_ADD("ddl", "create_partition", K(ret), "partition", batch_arg.args_.at(0).partition_key_, K(update_pt_cost), K(cost), K(ret)); } else { SERVER_EVENT_ADD("ddl", "batch_create_partition", K(ret), "partition_count", batch_arg.args_.count(), K(cost), K(update_pt_cost), "total_partition_cnt", partition_count, K(ret)); } return ret; } /* * input: * data_version * first: arg of first create_partition * batch_arg: arg of batch create partition * */ int ObService::sync_report_replica_info( const int64_t data_version, const ObCreatePartitionArg& first, const obrpc::ObCreatePartitionBatchArg& batch_arg) { int ret = OB_SUCCESS; if (OB_ISNULL(gctx_.sql_proxy_)) { ret = OB_ERR_UNEXPECTED; LOG_WARN("sql proxy is null", KR(ret)); } else { ObNormalPartitionTableProxy pt_proxy(*gctx_.sql_proxy_); ObTimeoutCtx timeout_ctx; int64_t timeout_us = THIS_WORKER.get_timeout_remain(); const int64_t MAX_EXECUTE_TIMEOUT_US = 2L * 100 * 1000; // 200ms // The meaning of REMAIN_TIME: the reserved time for statement and the return of RPC // after Synchronous reporting meta table const int64_t REMAIN_TIME = 1L * 100 * 1000; // 100ms int64_t stmt_timeout = MAX_EXECUTE_TIMEOUT_US; if (INT64_MAX != THIS_WORKER.get_timeout_ts()) { // if timeout is setted, the longest waiting time is 200ms except for the reserved time of 100ms timeout_us = std::min(timeout_us - REMAIN_TIME, stmt_timeout); } else { // just wait for 200ms, if timeout is not setted timeout_us = stmt_timeout; } if (timeout_us <= 0) { // nothing todo // At this time, the remaining time of the statement is less than the reserved value, // and it is no need to synchronous report } else if (OB_FAIL(timeout_ctx.set_timeout(timeout_us))) { LOG_WARN("failed to set timeout", KR(ret), K(timeout_us)); } else if (OB_FAIL(pt_proxy.update_fresh_replica_version( gctx_.self_addr_, GCONF.mysql_port, batch_arg.args_, data_version))) { LOG_WARN("update fresh replica version failed", KR(ret), K(gctx_.self_addr_), "data_version", first.memstore_version_ - 1); } if (OB_SUCC(ret)) { if (timeout_us <= 0) { ret = OB_TIMEOUT; } } } return ret; } int ObService::check_unique_index_request(const obrpc::ObCheckUniqueIndexRequestArg& arg) { int ret = OB_SUCCESS; if (OB_UNLIKELY(!inited_)) { ret = OB_NOT_INIT; LOG_WARN("ObService has not been inited", K(ret)); } else if (OB_UNLIKELY(!arg.is_valid())) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid arguments", K(ret), K(arg)); } else if (OB_ISNULL(gctx_.par_ser_)) { ret = OB_ERR_UNEXPECTED; LOG_WARN("error unexpected, partition service must not be NULL", K(ret)); } else { // For the time being, write a fake unique check, and remove it after the column check is completed obrpc::ObCheckUniqueIndexResponseArg response_arg; ObAddr rs_addr; response_arg.pkey_ = arg.pkey_; response_arg.index_id_ = arg.index_id_; response_arg.ret_code_ = OB_SUCCESS; response_arg.is_valid_ = true; if (OB_ISNULL(GCTX.rs_rpc_proxy_) || OB_ISNULL(GCTX.rs_mgr_)) { ret = OB_ERR_SYS; LOG_WARN("innner system error, rootserver rpc proxy or rs mgr must not be NULL", K(ret), K(GCTX)); } else if (OB_FAIL(GCTX.rs_mgr_->get_master_root_server(rs_addr))) { LOG_WARN("fail to get rootservice address", K(ret)); } else if (OB_FAIL(GCTX.rs_rpc_proxy_->to(rs_addr).check_unique_index_response(response_arg))) { LOG_WARN("fail to check unique index response", K(ret), K(arg)); } } return ret; } int ObService::calc_column_checksum_request(const obrpc::ObCalcColumnChecksumRequestArg& arg) { int ret = OB_SUCCESS; if (OB_UNLIKELY(!inited_)) { ret = OB_NOT_INIT; LOG_WARN("ObService has not been inited", K(ret)); } else if (OB_UNLIKELY(!arg.is_valid())) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid arguments", K(ret), K(arg)); } else if (OB_ISNULL(gctx_.par_ser_)) { ret = OB_ERR_UNEXPECTED; LOG_WARN("error unexpected, partition service must not be NULL", K(ret)); } else if (OB_FAIL(gctx_.par_ser_->calc_column_checksum( arg.pkey_, arg.index_id_, arg.schema_version_, arg.execution_id_, arg.snapshot_version_))) { LOG_WARN("fail to do calculate column checksum request", K(ret)); } return ret; } int ObService::check_single_replica_major_sstable_exist(const obrpc::ObCheckSingleReplicaMajorSSTableExistArg& arg) { int ret = OB_SUCCESS; if (OB_UNLIKELY(!inited_)) { ret = OB_NOT_INIT; LOG_WARN("ObService has not been inited", K(ret)); } else if (OB_UNLIKELY(!arg.is_valid())) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid arguments", K(ret), K(arg)); } else if (OB_ISNULL(gctx_.par_ser_)) { ret = OB_ERR_UNEXPECTED; LOG_WARN("error unexpected, partition service must not be NULL", K(ret)); } else if (OB_FAIL(gctx_.par_ser_->check_single_replica_major_sstable_exist(arg.pkey_, arg.index_id_))) { LOG_WARN("fail to check self major sstable exist", K(ret), K(arg)); } return ret; } int ObService::check_single_replica_major_sstable_exist( const obrpc::ObCheckSingleReplicaMajorSSTableExistArg& arg, obrpc::ObCheckSingleReplicaMajorSSTableExistResult& res) { int ret = OB_SUCCESS; if (OB_UNLIKELY(!inited_)) { ret = OB_NOT_INIT; LOG_WARN("ObService has not been inited", K(ret)); } else if (OB_UNLIKELY(!arg.is_valid())) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid arguments", K(ret), K(arg)); } else if (OB_ISNULL(gctx_.par_ser_)) { ret = OB_ERR_UNEXPECTED; LOG_WARN("error unexpected, partition service must not be NULL", K(ret)); } else if (OB_FAIL( gctx_.par_ser_->check_single_replica_major_sstable_exist(arg.pkey_, arg.index_id_, res.timestamp_))) { LOG_WARN("fail to check self major sstable exist", K(ret), K(arg)); } return ret; } int ObService::check_all_replica_major_sstable_exist(const obrpc::ObCheckAllReplicaMajorSSTableExistArg& arg) { int ret = OB_SUCCESS; if (OB_UNLIKELY(!inited_)) { ret = OB_NOT_INIT; LOG_WARN("ObService has not been inited", K(ret)); } else if (OB_UNLIKELY(!arg.is_valid())) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid arguments", K(ret), K(arg)); } else if (OB_ISNULL(gctx_.par_ser_)) { ret = OB_ERR_UNEXPECTED; LOG_WARN("error unexpected, partition service must not be NULL", K(ret)); } else if (OB_FAIL(gctx_.par_ser_->check_all_replica_major_sstable_exist(arg.pkey_, arg.index_id_))) { LOG_WARN("fail to check self major sstable exist", K(ret), K(arg)); } return ret; } int ObService::check_all_replica_major_sstable_exist( const obrpc::ObCheckAllReplicaMajorSSTableExistArg& arg, obrpc::ObCheckAllReplicaMajorSSTableExistResult& res) { int ret = OB_SUCCESS; if (OB_UNLIKELY(!inited_)) { ret = OB_NOT_INIT; LOG_WARN("ObService has not been inited", K(ret)); } else if (OB_UNLIKELY(!arg.is_valid())) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid arguments", K(ret), K(arg)); } else if (OB_ISNULL(gctx_.par_ser_)) { ret = OB_ERR_UNEXPECTED; LOG_WARN("error unexpected, partition service must not be NULL", K(ret)); } else if (OB_FAIL( gctx_.par_ser_->check_all_replica_major_sstable_exist(arg.pkey_, arg.index_id_, res.max_timestamp_))) { LOG_WARN("fail to check self major sstable exist", K(ret), K(arg)); } return ret; } int ObService::fetch_root_partition(share::ObPartitionReplica& replica) { int ret = OB_SUCCESS; if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else { const ObPartitionKey partition_key(combine_id(OB_SYS_TENANT_ID, OB_ALL_CORE_TABLE_TID), ObIPartitionTable::ALL_CORE_TABLE_PARTITION_ID, ObIPartitionTable::ALL_CORE_TABLE_PARTITION_NUM); if (OB_FAIL(fill_partition_replica(partition_key, replica))) { LOG_WARN("fetch_root_partition failed", K(ret), K(partition_key)); } else { LOG_INFO("fetch root partition succeed", K(replica)); } } return ret; } int ObService::add_replica(const obrpc::ObAddReplicaArg& arg, const share::ObTaskId& task_id) { LOG_INFO("receive add replica request", K(arg)); int ret = OB_SUCCESS; if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("ObService not init", K(ret)); } else if (!arg.is_valid()) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid arg", K(arg), K(ret)); } else if (OB_FAIL(gctx_.par_ser_->add_replica(arg, task_id))) { LOG_WARN("add replica fail", K(arg), K(ret)); } else { LOG_INFO("add replica successfully", K(arg)); } SERVER_EVENT_ADD("ddl", "add_replica", K(ret), "partition", arg.key_); return ret; } int ObService::remove_member(const obrpc::ObMemberChangeArg& arg) { LOG_INFO("receive remove member request", K(arg)); obrpc::ObMCLogRpcInfo mc_log_info; int ret = OB_SUCCESS; if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not inited", K(ret)); } else if (!arg.is_valid()) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid arg", K(arg), K(ret)); } else if (OB_FAIL(gctx_.par_ser_->remove_replica_mc(arg, mc_log_info))) { LOG_WARN("remove member failed", K(ret), K(arg)); } else { LOG_INFO("remove member success", K(arg)); } SERVER_EVENT_ADD("ddl", "remove_member", K(ret), "partition", arg.key_, "member", arg.member_); return ret; } int ObService::modify_quorum(const obrpc::ObModifyQuorumArg& arg) { LOG_INFO("receive modify quorum request", K(arg)); int ret = OB_SUCCESS; ObMCLogRpcInfo mc_log_info; if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not inited", K(ret)); } else if (!arg.is_valid()) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid arg", K(arg), K(ret)); } else if (OB_ISNULL(gctx_.par_ser_)) { ret = OB_ERR_UNEXPECTED; LOG_WARN("partition service is null", K(ret)); } else if (OB_FAIL(gctx_.par_ser_->change_quorum_mc(arg, mc_log_info))) { LOG_WARN("modify quorum failed", K(ret), K(arg)); } else { LOG_INFO("modify quorum success", K(arg)); } SERVER_EVENT_ADD( "ddl", "change_quorum", K(ret), "partition", arg.key_, "quorum", arg.quorum_, "orig_quorum", arg.quorum_); return ret; } int ObService::add_replica_batch(const obrpc::ObAddReplicaBatchArg& arg) { int ret = OB_SUCCESS; LOG_INFO("receive add_replica_batch request", K(arg)); if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("ObService not init", K(ret)); } else if (!arg.is_valid()) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid arg", K(arg), K(ret)); } else if (OB_FAIL(gctx_.par_ser_->batch_add_replica(arg, arg.task_id_))) { LOG_WARN("fail to add replica batch", K(ret), K(arg)); } else { LOG_INFO("add_replica_batch successfully", K(arg)); } return ret; } int ObService::remove_non_paxos_replica_batch( const obrpc::ObRemoveNonPaxosReplicaBatchArg& arg, obrpc::ObRemoveNonPaxosReplicaBatchResult& result) { int ret = OB_SUCCESS; const int64_t start = ObTimeUtility::current_time(); LOG_INFO("receive remove_non_paxos_replica_batch request", "arg_cnt", arg.arg_array_.count()); if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("ObService not init", K(ret)); } else if (!arg.is_valid()) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid arg", K(arg), K(ret)); } else if (OB_ISNULL(gctx_.par_ser_)) { ret = OB_ERR_UNEXPECTED; LOG_WARN("partition service is null", K(ret)); } else if (OB_FAIL(gctx_.par_ser_->batch_remove_non_paxos_replica(arg, result))) { LOG_WARN("remove_non_paxos_replica_batch fail", K(ret), K(arg)); } const int64_t cost = ObTimeUtility::current_time() - start; LOG_INFO("batch remove non paxos repilca cost", "arg_cnt", arg.arg_array_.count(), K(cost)); SERVER_EVENT_ADD("ddl", "batch_remove_non_paxos_replica", K(ret), K(cost), "partition_cnt", arg.arg_array_.count(), "result_cnt", result.return_array_.count(), "ret", ret); return ret; } int ObService::remove_member_batch(const obrpc::ObMemberChangeBatchArg& arg, obrpc::ObMemberChangeBatchResult& result) { int ret = OB_SUCCESS; const int64_t start = ObTimeUtility::current_time(); LOG_INFO("receive remove_member_batch request", "arg_cnt", arg.arg_array_.count()); if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("ObService not init", K(ret)); } else if (!arg.is_valid()) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid arg", K(arg), K(ret)); } else if (OB_ISNULL(gctx_.par_ser_)) { ret = OB_ERR_UNEXPECTED; LOG_WARN("partition service is null", K(ret)); } else if (OB_FAIL(gctx_.par_ser_->batch_remove_replica_mc(arg, result))) { LOG_WARN("batch remove member failed", K(ret), "arg_cnt", arg.arg_array_.count()); } const int64_t cost = ObTimeUtility::current_time() - start; LOG_INFO("batch remove member cost", "arg_cnt", arg.arg_array_.count(), K(cost)); SERVER_EVENT_ADD("ddl", "batch_remove_member", K(ret), K(cost), "partition_cnt", arg.arg_array_.count(), "result_cnt", result.return_array_.count(), "ret", ret); return ret; } int ObService::modify_quorum_batch(const obrpc::ObModifyQuorumBatchArg& arg, obrpc::ObModifyQuorumBatchResult& result) { int ret = OB_SUCCESS; const int64_t start = ObTimeUtility::current_time(); LOG_INFO("receive modify_quorum_batch request", "arg_cnt", arg.arg_array_.count()); if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("ObService not init", K(ret)); } else if (!arg.is_valid()) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid arg", K(arg), K(ret)); } else if (OB_ISNULL(gctx_.par_ser_)) { ret = OB_ERR_UNEXPECTED; LOG_WARN("partition service is null", K(ret)); } else { for (int64_t i = 0; OB_SUCC(ret) && i < arg.arg_array_.count(); i++) { const ObModifyQuorumArg& mc_arg = arg.arg_array_.at(i); if (OB_FAIL(result.return_array_.push_back(OB_ERROR))) { LOG_WARN("fail to add ret", K(ret)); } else { ObMCLogRpcInfo mc_log_info; int tmp_ret = OB_SUCCESS; if (OB_SUCCESS != (tmp_ret = gctx_.par_ser_->change_quorum_mc(mc_arg, mc_log_info))) { LOG_WARN("fail to change quorum", K(ret), K(mc_arg)); } result.return_array_.at(i) = tmp_ret; } } } const int64_t cost = ObTimeUtility::current_time() - start; LOG_INFO("batch modify quorum cost", "arg_cnt", arg.arg_array_.count(), K(cost)); SERVER_EVENT_ADD("ddl", "batch_modify_quorum", K(ret), K(cost), "partition_cnt", arg.arg_array_.count(), "result_cnt", result.return_array_.count(), "ret", ret); return ret; } int ObService::migrate_replica_batch(const obrpc::ObMigrateReplicaBatchArg& arg) { int ret = OB_SUCCESS; LOG_INFO("receive migrate replica batch request", K(arg)); if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("ObService not init", K(ret)); } else if (!arg.is_valid()) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid arg", K(arg), K(ret)); } else if (OB_FAIL(gctx_.par_ser_->migrate_replica_batch(arg))) { LOG_WARN("add migrate batch fail", K(arg), K(ret)); } else { LOG_INFO("add migrate batch successfully", K(arg)); } return ret; } int ObService::standby_cutdata_batch_task(const obrpc::ObStandbyCutDataBatchTaskArg& arg) { int ret = OB_SUCCESS; LOG_INFO("receive standby_cutdata_batch_task request", K(arg)); return ret; } int ObService::rebuild_replica_batch(const obrpc::ObRebuildReplicaBatchArg& arg) { int ret = OB_SUCCESS; LOG_INFO("receive rebuild_replica_batch request", K(arg)); if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("ObService not init", K(ret)); } else if (!arg.is_valid()) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid arg", K(arg), K(ret)); } else if (1 != arg.arg_array_.count()) { ret = OB_NOT_SUPPORTED; LOG_ERROR("rebuild_replica_batch is not supported yet", K(ret), K(arg)); } else if (OB_FAIL(rebuild_replica(arg.arg_array_.at(0), arg.task_id_))) { LOG_WARN("rebuild_replica_batch fail", K(ret), K(arg)); } else { LOG_INFO("rebuild_replica_batch successfully", K(arg)); } return ret; } int ObService::backup_replica_batch(const obrpc::ObBackupBatchArg& arg) { int ret = OB_SUCCESS; LOG_INFO("receive backup replica batch request", K(arg)); if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("ObService not init", K(ret)); } else if (!arg.is_valid()) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid arg", K(arg), K(ret)); } else if (OB_FAIL(gctx_.par_ser_->backup_replica_batch(arg))) { LOG_WARN("add migrate batch fail", K(arg), K(ret)); } else { LOG_INFO("backup replica batch successfully", K(arg)); } return ret; } int ObService::validate_backup_batch(const obrpc::ObValidateBatchArg& arg) { int ret = OB_SUCCESS; LOG_INFO("receive validate backup batch request", K(arg)); ; if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("ObService not init", K(ret)); } else if (!arg.is_valid()) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid argument", K(ret), K(arg)); } else if (OB_FAIL(gctx_.par_ser_->validate_backup_batch(arg))) { LOG_WARN("validate backup batch fail", K(ret), K(arg)); } else { LOG_INFO("validate backup batch successfully", K(arg)); } return ret; } int ObService::check_sys_task_exist(const share::ObTaskId& arg, bool& res) { int ret = OB_SUCCESS; if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("ObService not init", K(ret)); } else if (arg.is_invalid()) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid arg", K(arg), K(ret)); } else if (OB_FAIL(SYS_TASK_STATUS_MGR.task_exist(arg, res))) { LOG_WARN("failed to check task exist", K(ret), K(arg)); } return ret; } int ObService::check_migrate_task_exist(const share::ObTaskId& arg, bool& res) { int ret = OB_SUCCESS; if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("ObService not init", K(ret)); } else if (arg.is_invalid()) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid arg", K(arg), K(ret)); } else if (OB_FAIL(ObPartGroupMigrator::get_instance().task_exist(arg, res))) { LOG_WARN("failed to check migrate task exist", K(ret), K(arg)); } return ret; } int ObService::change_replica_batch(const obrpc::ObChangeReplicaBatchArg& arg) { int ret = OB_SUCCESS; LOG_INFO("receive change_replica_batch request", K(arg)); if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("ObService not init", K(ret)); } else if (!arg.is_valid()) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid arg", K(arg), K(ret)); } else if (OB_FAIL(gctx_.par_ser_->batch_change_replica(arg))) { LOG_WARN("change_replica_batch fail", K(ret), K(arg)); } else { LOG_INFO("change_replica_batch successfully", K(arg)); } if (arg.arg_array_.count() > 0) { SERVER_EVENT_ADD("ddl", "batch_change_replica", K(ret), "task_id", arg.task_id_, "count", arg.arg_array_.count(), "from", arg.arg_array_.at(0).src_, "to", arg.arg_array_.at(0).dst_, "quorum", arg.arg_array_.at(0).quorum_); } return ret; } int ObService::report_replica(const obrpc::ObReportSingleReplicaArg& arg) { int ret = OB_SUCCESS; LOG_INFO("receive report replica request", K(arg.partition_key_)); const bool need_report_checksum = false; const bool with_role_report = true; ObPartitionArray pkeys; if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else if (!arg.partition_key_.is_valid()) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid argument", K(arg.partition_key_), K(ret)); } else { ObIPartitionGroup* pg = NULL; ObIPartitionGroupGuard guard; // If the replica is pg, it is need to update all partitions under pg if (OB_FAIL(gctx_.par_ser_->get_partition(arg.partition_key_, guard))) { if (OB_ENTRY_NOT_EXIST != ret && OB_PARTITION_NOT_EXIST != ret) { LOG_WARN("get partition failed", K(ret), K(arg)); } else { // the partition was deleted, no need to report; ret = OB_SUCCESS; } } else if (NULL == (pg = guard.get_partition_group())) { ret = OB_ERR_UNEXPECTED; LOG_WARN("pg is null", KP(pg), K(ret)); } else if (OB_FAIL(pg->get_all_pg_partition_keys(pkeys))) { LOG_WARN("get all pg partition keys error", K(arg)); } else { // do nothing } } if (OB_SUCC(ret)) { if (OB_FAIL(submit_pt_update_task(arg.partition_key_, need_report_checksum, with_role_report))) { LOG_WARN("async_update failed", K(arg.partition_key_), K(ret)); } // update partition meta table, ignore failed submit_pg_pt_update_task(pkeys); } SERVER_EVENT_ADD("storage", "report_replica", K(ret), "partition", arg.partition_key_); return ret; } int ObService::migrate_replica(const obrpc::ObMigrateReplicaArg& arg, const share::ObTaskId& task_id) { LOG_INFO("receive migrate replica request", K(arg)); int ret = OB_SUCCESS; if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else if (!arg.is_valid()) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid arg", K(arg), K(ret)); } else if (OB_FAIL(gctx_.par_ser_->migrate_replica(arg, task_id))) { LOG_WARN("migrate replica fail", K(arg), K(ret)); } else { LOG_INFO("migrate replica successfully", K(arg)); } SERVER_EVENT_ADD("ddl", "migrate_replica", K(ret), "partition", arg.key_, "from", arg.src_, "to", arg.dst_); return ret; } int ObService::minor_freeze(const obrpc::ObMinorFreezeArg& arg, obrpc::Int64& result) { int ret = OB_SUCCESS; const int64_t start_ts = ObTimeUtility::current_time(); LOG_INFO("receive minor freeze request", K(arg), "is_pkey_valid", arg.partition_key_.is_valid()); if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else if (!arg.is_valid()) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid arg", K(arg), K(ret)); } else if (OB_ISNULL(gctx_.par_ser_)) { ret = OB_ERR_UNEXPECTED; LOG_WARN("partition service shouldn't be null in gctx", K(ret)); } else { if (arg.partition_key_.is_valid()) { if (OB_FAIL(gctx_.par_ser_->minor_freeze(arg.partition_key_))) { LOG_WARN("fail to do minor freeze, ", "pkey", arg.partition_key_, K(ret)); } } else if (arg.tenant_ids_.count() > 0) { // minor freeze with tenants for (int i = 0; i < arg.tenant_ids_.count(); ++i) { int tmp_ret = OB_SUCCESS; uint64_t tenant_id = arg.tenant_ids_.at(i); if (OB_UNLIKELY(OB_SUCCESS != (tmp_ret = gctx_.par_ser_->minor_freeze(tenant_id)))) { LOG_WARN("fail to do minor freeze", K(tenant_id), K(tmp_ret)); if (OB_SUCC(ret)) { ret = tmp_ret; } } } } else { if (OB_FAIL(gctx_.par_ser_->minor_freeze(OB_INVALID_TENANT_ID))) { LOG_WARN("fail to minor freeze all", K(ret)); } } } result = ret; const int64_t cost_ts = ObTimeUtility::current_time() - start_ts; LOG_INFO("finish minor freeze request", K(ret), K(arg), K(cost_ts)); return ret; } int ObService::remove_non_paxos_replica(const obrpc::ObRemoveNonPaxosReplicaArg& arg) { int ret = OB_SUCCESS; LOG_INFO("receive remove replica request", K(arg)); if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("ObService not init", K(ret)); } else if (!arg.is_valid()) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid arg", K(arg), K(ret)); } else if (OB_FAIL(gctx_.par_ser_->remove_replica(arg.key_, arg.dst_))) { LOG_WARN("remove replica fail", K(arg), K(ret)); } else { LOG_INFO("remove replica successfully", K(arg)); } SERVER_EVENT_ADD("ddl", "remove_replica", K(ret), "partition", arg.key_); return ret; } int ObService::rebuild_replica(const obrpc::ObRebuildReplicaArg& arg, const share::ObTaskId& task_id) { LOG_INFO("receive rebuild replica request", K(arg)); int ret = OB_SUCCESS; if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else if (!arg.is_valid()) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid arg", K(arg), K(ret)); } else if (OB_FAIL(gctx_.par_ser_->rebuild_replica(arg, task_id))) { LOG_WARN("rebuild replica fail", K(arg), K(ret)); } else { LOG_INFO("rebuild replica successfully", K(arg)); } SERVER_EVENT_ADD("ddl", "rebuild_replica", K(ret), "partition", arg.key_, "from", arg.src_, "to", arg.dst_); return ret; } int ObService::restore_replica(const obrpc::ObRestoreReplicaArg& arg, const share::ObTaskId& task_id) { LOG_INFO("receive restore replica request", K(arg)); int ret = OB_SUCCESS; if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else if (!arg.is_valid()) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid arg", K(arg), K(ret)); } else if (OB_FAIL(gctx_.par_ser_->restore_replica(arg, task_id))) { LOG_WARN("failed to restore_replica", K(arg), K(ret)); } else { LOG_INFO("restore replica successfully", K(arg)); } SERVER_EVENT_ADD("restore", "restore_replica", K(ret), "partition", arg.key_, "from", arg.src_, "to", arg.dst_); return ret; } int ObService::physical_restore_replica(const obrpc::ObPhyRestoreReplicaArg& arg, const share::ObTaskId& task_id) { LOG_INFO("receive physical restore replica request", K(arg)); int ret = OB_SUCCESS; if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else if (!arg.is_valid()) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid arg", K(arg), K(ret)); } else if (OB_FAIL(gctx_.par_ser_->physical_restore_replica(arg, task_id))) { LOG_WARN("failed to restore_replica", K(arg), K(ret)); } else { LOG_INFO("physical restore replica success", K(arg)); } SERVER_EVENT_ADD( "restore", "physical_restore_replica", K(ret), "partition", arg.key_, "from", arg.src_, "to", arg.dst_); return ret; } int ObService::get_tenant_log_archive_status( const share::ObGetTenantLogArchiveStatusArg& arg, share::ObTenantLogArchiveStatusWrapper& result) { LOG_INFO("receive get_tenant_log_archive_status request", K(arg)); int ret = OB_SUCCESS; if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else if (!arg.is_valid()) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid arg", K(arg), K(ret)); } else if (OB_FAIL(gctx_.par_ser_->get_tenant_log_archive_status(arg, result))) { LOG_WARN("failed to restore_replica", K(arg), K(ret)); } else { LOG_INFO("finish get_tenant_log_archive_status request", K(arg), K(result)); } return ret; } int ObService::copy_sstable_batch(const obrpc::ObCopySSTableBatchArg& arg) { LOG_INFO("receive copy sstable batch request", K(arg)); int ret = OB_SUCCESS; if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else if (!arg.is_valid()) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid arg", K(arg), K(ret)); } else if (OB_COPY_SSTABLE_TYPE_INVALID == arg.type_) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid copy sstable type", K(ret), K(arg)); } else if (1 != arg.arg_array_.count()) { ret = OB_NOT_SUPPORTED; LOG_ERROR("copy_sstable_batch is not supported yet", K(ret), K(arg)); } else { switch (arg.type_) { case OB_COPY_SSTABLE_TYPE_LOCAL_INDEX: { for (int64_t i = 0; OB_SUCC(ret) && i < arg.arg_array_.count(); ++i) { const obrpc::ObCopySSTableArg& single_arg = arg.arg_array_.at(i); if (OB_FAIL(gctx_.par_ser_->copy_local_index(single_arg, arg.task_id_))) { LOG_WARN("fail to copy local index", K(ret), K(single_arg)); } } break; } case OB_COPY_SSTABLE_TYPE_GLOBAL_INDEX: { for (int64_t i = 0; OB_SUCC(ret) && i < arg.arg_array_.count(); ++i) { const obrpc::ObCopySSTableArg& single_arg = arg.arg_array_.at(i); if (OB_FAIL(gctx_.par_ser_->copy_global_index(single_arg, arg.task_id_))) { LOG_WARN("fail to copy global index", K(ret), K(single_arg)); } } break; } case OB_COPY_SSTABLE_TYPE_RESTORE_FOLLOWER: { if (OB_FAIL(gctx_.par_ser_->restore_follower_replica(arg))) { if (OB_RESTORE_PARTITION_IS_COMPELETE == ret) { LOG_INFO("restore follower replica is already finished", K(ret)); } else { LOG_WARN("failed to restore follower", K(ret), K(arg)); } } break; } default: { ret = OB_NOT_SUPPORTED; LOG_WARN("invalid copy sstable rpc type", K(ret), "type", arg.type_); } } } if (arg.arg_array_.count() > 0) { SERVER_EVENT_ADD("ddl", "copy_sstable_batch", K(ret), "partition", arg.arg_array_.at(0).key_, "src", arg.arg_array_.at(0).src_, "dst", arg.arg_array_.at(0).dst_); } else { SERVER_EVENT_ADD("ddl", "copy_sstable_batch", K(ret)); } return ret; } int ObService::change_replica(const obrpc::ObChangeReplicaArg& arg, const share::ObTaskId& task_id) { int ret = OB_SUCCESS; LOG_INFO("receive change replica info request", K(arg)); if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else if (!arg.is_valid()) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid arg", K(arg), K(ret)); } else if (OB_FAIL(gctx_.par_ser_->change_replica(arg, task_id))) { LOG_WARN("change replica info fail", K(arg), K(ret)); } else { LOG_INFO("change replica info successfully", K(arg)); } SERVER_EVENT_ADD( "ddl", "change_replica", K(ret), "partition", arg.key_, "from", arg.src_, "to", arg.dst_, "quorum", arg.quorum_); return ret; } int ObService::check_ctx_create_timestamp_elapsed( const obrpc::ObCheckCtxCreateTimestampElapsedArg& arg, obrpc::ObCheckCtxCreateTimestampElapsedResult& result) { int ret = OB_SUCCESS; ObRole role; LOG_INFO("receive get checksum cal snapshot", K(arg)); if (OB_UNLIKELY(!inited_)) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else if (OB_UNLIKELY(!arg.is_valid())) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid argument", K(ret), K(arg)); } else if (OB_UNLIKELY(NULL == gctx_.par_ser_)) { ret = OB_ERR_UNEXPECTED; LOG_WARN("partition service shouldn't be null in gctx", K(ret)); } else if (OB_FAIL(gctx_.par_ser_->get_role(arg.pkey_, role))) { LOG_WARN("fail to get role", K(ret)); } else if (!is_strong_leader(role)) { ret = OB_NOT_MASTER; } else if (OB_FAIL(gctx_.par_ser_->check_ctx_create_timestamp_elapsed(arg.pkey_, arg.sstable_exist_ts_))) { if (OB_EAGAIN != ret) { LOG_WARN("check ctx create timestamp failed", K(ret)); } } else { transaction::ObTransService* txs = gctx_.par_ser_->get_trans_service(); if (OB_UNLIKELY(NULL == txs)) { ret = OB_ERR_UNEXPECTED; LOG_WARN("txs ptr is null", K(ret)); } else if (OB_FAIL(txs->get_publish_version(arg.pkey_, result.snapshot_))) { LOG_WARN("fail to get publish version", K(ret)); } else if (OB_FAIL(gctx_.par_ser_->get_role(arg.pkey_, role))) { LOG_WARN("fail to get role", K(ret)); } else if (!is_strong_leader(role)) { ret = OB_NOT_MASTER; } } return ret; } int ObService::check_schema_version_elapsed( const obrpc::ObCheckSchemaVersionElapsedArg& arg, obrpc::ObCheckSchemaVersionElapsedResult& result) { int ret = OB_SUCCESS; LOG_INFO("receive check schema version elapsed", K(arg)); if (OB_UNLIKELY(!inited_)) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else if (OB_UNLIKELY(!arg.is_valid())) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid argument", K(ret), K(arg)); } else if (OB_UNLIKELY(NULL == gctx_.par_ser_)) { ret = OB_ERR_UNEXPECTED; LOG_WARN("partition service shouldn't be null in gctx", K(ret)); } else if (OB_FAIL(gctx_.par_ser_->check_schema_version_elapsed( arg.pkey_, arg.schema_version_, 0 /*index_id*/, result.snapshot_))) { LOG_WARN("fail to do check schema version elapsed", K(ret)); } else { } // no more to do return ret; } int ObService::get_member_list(const ObPartitionKey& partition_key, obrpc::ObServerList& members) { // TODO(): add logic UNUSED(partition_key); UNUSED(members); return OB_SUCCESS; } int ObService::switch_leader(const obrpc::ObSwitchLeaderArg& arg) { int ret = OB_SUCCESS; LOG_INFO("receive switch leader request", K(arg)); if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else if (!arg.is_valid()) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid arg", K(arg), K(ret)); } else if (OB_FAIL(gctx_.par_ser_->change_leader(arg.partition_key_, arg.leader_addr_))) { LOG_WARN("change_leader failed", K(arg), K(ret)); } SERVER_EVENT_ADD("election", "switch_leader", K(ret), "partition", arg.partition_key_, "leader", arg.leader_addr_); return ret; } int ObService::batch_switch_rs_leader(const ObAddr& arg) { int ret = OB_SUCCESS; LOG_INFO("receive batch switch rs leader request", K(arg)); int64_t start_timestamp = ObTimeUtility::current_time(); if (OB_UNLIKELY(!inited_)) { ret = OB_NOT_INIT; LOG_WARN("not init", KR(ret)); } else if (OB_ISNULL(gctx_.par_ser_)) { ret = OB_ERR_UNEXPECTED; LOG_WARN("gctx par_ser is NULL", K(arg)); } else if (!arg.is_valid()) { if (OB_FAIL(gctx_.par_ser_->auto_batch_change_rs_leader())) { LOG_WARN("fail to auto batch change rs leader", KR(ret)); } } else if (OB_FAIL(gctx_.par_ser_->batch_change_rs_leader(arg))) { LOG_WARN("fail to batch change rs leader", K(arg), KR(ret)); } int64_t cost = ObTimeUtility::current_time() - start_timestamp; SERVER_EVENT_ADD("election", "batch_switch_rs_leader", K(ret), "leader", arg, K(cost)); return ret; } int ObService::switch_leader_list(const obrpc::ObSwitchLeaderListArg& arg) { int ret = OB_SUCCESS; LOG_INFO("receive switch leader list request", K(arg)); int64_t start_timestamp = ObTimeUtility::current_time(); if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else if (!arg.is_valid()) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid arg", K(arg), K(ret)); } else { for (int64_t i = 0; OB_SUCC(ret) && i < arg.partition_key_list_.count(); ++i) { if (OB_FAIL(gctx_.par_ser_->change_leader(arg.partition_key_list_.at(i), arg.leader_addr_))) { LOG_WARN("change_leader failed", K(ret), K(i), K(arg.partition_key_list_.at(i)), K(arg.leader_addr_)); } } } int64_t cost = ObTimeUtility::current_time() - start_timestamp; if (arg.partition_key_list_.count() == 1) { SERVER_EVENT_ADD("election", "switch_leader", K(ret), "partition", arg.partition_key_list_.at(0), "leader", arg.leader_addr_, K(cost)); } else { SERVER_EVENT_ADD("election", "batch_switch_leader", K(ret), "partition_count", arg.partition_key_list_.count(), "leader", arg.leader_addr_, K(cost)); } return ret; } int ObService::get_leader_candidates_v2(const ObGetLeaderCandidatesV2Arg& arg, ObGetLeaderCandidatesResult& result) { // It is used to obtain valid_candidates in batch after 2.0 int ret = OB_SUCCESS; ObSArray candidate_list_array; ObSArray candidate_status_array; if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else if (!arg.is_valid()) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid arg", K(arg), K(ret)); } else if (OB_FAIL(gctx_.par_ser_->get_dst_candidates_array( arg.partitions_, arg.prep_candidates_, candidate_list_array, candidate_status_array))) { LOG_WARN("get_dst_candidates_array failed", K(ret)); } else if (candidate_list_array.count() != candidate_status_array.count()) { ret = OB_ERR_UNEXPECTED; LOG_WARN("array count not match", K(ret), "candidate_cnt", candidate_list_array.count(), "status_cnt", candidate_status_array.count()); } else { DEBUG_SYNC(OBSERVICE_GET_LEADER_CANDIDATES); for (int64_t i = 0; OB_SUCC(ret) && i < candidate_list_array.count(); ++i) { if (OB_FAIL(result.candidates_.push_back(candidate_list_array.at(i)))) { LOG_WARN("push_back failed", K(ret)); } else if (OB_FAIL(result.candidate_status_array_.push_back(candidate_status_array.at(i)))) { LOG_WARN("fail to push back", K(ret)); } } } return ret; } int ObService::get_leader_candidates(const ObGetLeaderCandidatesArg& arg, ObGetLeaderCandidatesResult& result) { int ret = OB_SUCCESS; if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else if (!arg.is_valid()) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid arg", K(arg), K(ret)); } else { DEBUG_SYNC(OBSERVICE_GET_LEADER_CANDIDATES); ObMemberList member_list; ObServerList candidate; ObAddr server; bool NEW_GET_LEADER_CANDIDATE = !(GET_MIN_CLUSTER_VERSION() < CLUSTER_VERSION_2000); FOREACH_CNT_X(partition, arg.partitions_, OB_SUCCESS == ret) { member_list.reset(); candidate.reset(); if (!NEW_GET_LEADER_CANDIDATE) { if (OB_FAIL(gctx_.par_ser_->get_dst_leader_candidate(*partition, member_list))) { LOG_WARN("get_dst_leader_candidate failed", "partition", *partition, K(ret)); } else { for (int64_t i = 0; OB_SUCC(ret) && i < member_list.get_member_number(); ++i) { server.reset(); if (OB_FAIL(member_list.get_server_by_index(i, server))) { LOG_WARN("get_server_by_index failed", K(i), K(ret)); } else if (OB_FAIL(candidate.push_back(server))) { LOG_WARN("push_back failed", K(ret)); } } if (OB_FAIL(ret)) { } else if (OB_FAIL(candidate.push_back(gctx_.self_addr_))) { LOG_WARN("push_back failed", K(ret)); } else if (OB_FAIL(result.candidates_.push_back(candidate))) { LOG_WARN("push_back failed", K(ret)); } } } else { ObGetLeaderCandidatesV2Arg this_arg; ObGetLeaderCandidatesResult this_result; if (OB_FAIL(this_arg.partitions_.push_back(*partition))) { LOG_WARN("fail to push back partition key", K(ret)); } else if (OB_FAIL(get_leader_member(*partition, this_arg.prep_candidates_))) { LOG_WARN("fail to get leader member", K(ret)); } else if (OB_FAIL(get_leader_candidates_v2(this_arg, this_result))) { LOG_WARN("fail to get leader candidates", K(ret)); } else if (this_result.candidates_.count() <= 0 || this_result.candidates_.count() > 1) { ret = OB_ERR_UNEXPECTED; LOG_WARN("result candidates count unexpected", K(ret)); } else if (this_result.candidate_status_array_.count() <= 0 || this_result.candidate_status_array_.count() > 1) { ret = OB_ERR_UNEXPECTED; LOG_WARN("result candidate status count unexpected", K(ret)); } else if (OB_FAIL(result.candidates_.push_back(this_result.candidates_.at(0)))) { LOG_WARN("fail to push back", K(ret)); } else if (OB_FAIL(result.candidate_status_array_.push_back(this_result.candidate_status_array_.at(0)))) { LOG_WARN("fail to push back", K(ret)); } else { } // no more to do } } } return ret; } int ObService::switch_schema(const obrpc::ObSwitchSchemaArg& arg) { int ret = OB_SUCCESS; LOG_INFO("start to switch schema", K(arg)); const ObRefreshSchemaInfo& schema_info = arg.schema_info_; const int64_t schema_version = schema_info.get_schema_version(); if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else if (schema_version <= 0) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid argument", K(schema_version), K(ret)); } else if (arg.force_refresh_) { ObSEArray tenant_ids; ObMultiVersionSchemaService* schema_service = gctx_.schema_service_; int64_t sys_schema_version = OB_INVALID_VERSION; if (schema_info.is_valid() && OB_INVALID_TENANT_ID != schema_info.get_tenant_id() && OB_SYS_TENANT_ID != schema_info.get_tenant_id()) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid schema info", K(ret), K(schema_info)); } else if (OB_FAIL(tenant_ids.push_back(OB_SYS_TENANT_ID))) { LOG_WARN("fail to push back tenant_id", K(ret)); } else if (OB_ISNULL(schema_service)) { ret = OB_ERR_UNEXPECTED; LOG_WARN("schema service is null", K(ret)); } else if (OB_FAIL(schema_service->refresh_and_add_schema(tenant_ids))) { LOG_WARN("fail to refresh schema", K(ret), K(schema_info)); } else if (schema_info.get_schema_version() <= 0) { // skip } else if (OB_FAIL(schema_service->get_tenant_refreshed_schema_version(OB_SYS_TENANT_ID, sys_schema_version))) { LOG_WARN("fail to get local sys schema_version", K(ret)); } else if (schema_info.get_schema_version() > sys_schema_version) { ret = OB_EAGAIN; LOG_WARN("schema is not new enough", K(ret), K(schema_info), K(sys_schema_version)); } } else if (!schema_info.is_valid()) { if (OB_FAIL(schema_updater_.try_reload_schema(schema_version))) { LOG_WARN("reload schema failed", K(schema_version), K(ret)); } else { LOG_INFO("switch schema version success", K(ret), K(schema_version)); } } else { if (OB_FAIL(schema_updater_.try_reload_schema(schema_info))) { LOG_WARN("reload schema failed", K(schema_info), K(ret)); } else { LOG_INFO("switch schema version success", K(ret), K(schema_info)); } } // SERVER_EVENT_ADD("schema", "switch_schema", K(ret), K(schema_info)); return ret; } int ObService::bootstrap(const obrpc::ObBootstrapArg& arg) { int ret = OB_SUCCESS; const int64_t timeout = 600 * 1000 * 1000LL; // 10 minutes const obrpc::ObServerInfoList& rs_list = arg.server_list_; LOG_INFO("bootstrap timeout", K(timeout), "worker_timeout_ts", THIS_WORKER.get_timeout_ts()); if (!inited_) { ret = OB_NOT_INIT; BOOTSTRAP_LOG(WARN, "not init", K(ret)); } else if (rs_list.count() <= 0) { ret = OB_INVALID_ARGUMENT; BOOTSTRAP_LOG(WARN, "rs_list is empty", K(rs_list), K(ret)); } else { ObPreBootstrap pre_bootstrap( *gctx_.srv_rpc_proxy_, rs_list, *gctx_.pt_operator_, *gctx_.config_, arg, *gctx_.rs_rpc_proxy_); ObAddr master_rs; bool server_empty = false; ObCheckServerEmptyArg new_arg; new_arg.mode_ = ObCheckServerEmptyArg::BOOTSTRAP; int64_t initial_frozen_version = 0; int64_t initial_schema_version = 0; obrpc::ObBootstrapArg tmp_arg; ObSArray frozen_status; ObSArray freeze_schemas; const bool wait_log_scan = true; if (OB_FAIL(check_server_empty(new_arg, wait_log_scan, server_empty))) { BOOTSTRAP_LOG(WARN, "check_server_empty failed", K(ret), K(new_arg)); } else if (!server_empty) { ret = OB_ERR_SYS; BOOTSTRAP_LOG(WARN, "observer is not empty", K(ret)); } else if (OB_FAIL(pre_bootstrap.prepare_bootstrap( master_rs, initial_frozen_version, initial_schema_version, frozen_status, freeze_schemas))) { BOOTSTRAP_LOG(ERROR, "failed to prepare boot strap", K(rs_list), K(ret)); } else if (OB_FAIL(tmp_arg.assign(arg))) { BOOTSTRAP_LOG(WARN, "fail to assign", KR(ret), K(arg)); } else if (OB_FAIL(tmp_arg.frozen_status_.assign(frozen_status))) { BOOTSTRAP_LOG(WARN, "failed to assign frozen status", KR(ret), K(frozen_status)); } else if (OB_FAIL(tmp_arg.freeze_schemas_.assign(freeze_schemas))) { BOOTSTRAP_LOG(WARN, "failed to assign freeze schemas", KR(ret), K(freeze_schemas)); } else { const ObCommonRpcProxy& rpc_proxy = *gctx_.rs_rpc_proxy_; bool boot_done = false; const int64_t MAX_RETRY_COUNT = 30; tmp_arg.initial_frozen_version_ = initial_frozen_version; tmp_arg.initial_schema_version_ = initial_schema_version; for (int i = 0; !boot_done && i < MAX_RETRY_COUNT; i++) { ret = OB_SUCCESS; int64_t rpc_timeout = timeout; if (INT64_MAX != THIS_WORKER.get_timeout_ts()) { rpc_timeout = max(rpc_timeout, THIS_WORKER.get_timeout_remain()); } if (OB_FAIL(rpc_proxy.to_addr(master_rs).timeout(rpc_timeout).execute_bootstrap(tmp_arg))) { if (OB_RS_NOT_MASTER == ret) { BOOTSTRAP_LOG( INFO, "master root service not ready", K(master_rs), "retry_count", i, K(rpc_timeout), K(ret)); SLEEP(1); } else { BOOTSTRAP_LOG(ERROR, "execute bootstrap fail", K(master_rs), K(rpc_timeout), K(ret)); break; } } else { boot_done = true; } } if (boot_done) { BOOTSTRAP_LOG(INFO, "succeed to do_boot_strap", K(rs_list), K(master_rs)); } } } return ret; } int ObService::check_deployment_mode_match(const obrpc::ObCheckDeploymentModeArg& arg, obrpc::Bool& match) { int ret = OB_SUCCESS; if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else { const bool single_zone_deployment_on = false; if (single_zone_deployment_on == arg.single_zone_deployment_on_) { match = true; } else { match = false; } } return ret; } int ObService::is_empty_server(const obrpc::ObCheckServerEmptyArg& arg, obrpc::Bool& is_empty) { int ret = OB_SUCCESS; if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else { bool server_empty = false; const bool wait_log_scan = (ObCheckServerEmptyArg::BOOTSTRAP == arg.mode_); if (OB_FAIL(check_server_empty(arg, wait_log_scan, server_empty))) { LOG_WARN("check_server_empty failed", K(ret), K(arg)); } else { is_empty = server_empty; } } return ret; } int ObService::get_partition_count(obrpc::ObGetPartitionCountResult& result) { int ret = OB_SUCCESS; result.reset(); if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not inited", K(ret)); } else if (OB_FAIL(gctx_.par_ser_->get_partition_count(result.partition_count_))) { LOG_WARN("failed to get partition count", K(ret)); } return ret; } int ObService::get_partition_stat(obrpc::ObPartitionStatList& partition_stat_list) { int ret = OB_SUCCESS; partition_stat_list.reuse(); if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else { ObIPartitionGroupIterator* partition_iter = gctx_.par_ser_->alloc_pg_iter(); if (NULL == partition_iter) { ret = OB_ALLOCATE_MEMORY_FAILED; LOG_ERROR("alloc_scan_iter failed", K(ret)); } else { ObPartitionStat partition_stat; ObIPartitionGroup* partition = NULL; while (OB_SUCC(ret)) { if (OB_FAIL(partition_iter->get_next(partition))) { if (OB_ITER_END != ret) { LOG_WARN("partition_iter get_next failed", K(ret)); } } else if (NULL == partition) { ret = OB_ERR_UNEXPECTED; LOG_WARN("partition is null", KP(partition), K(ret)); } else { partition_stat.reset(); partition_stat.partition_key_ = partition->get_partition_key(); partition_stat.stat_ = ObPartitionStat::WRITABLE; if (OB_FAIL(partition_stat_list.push_back(partition_stat))) { LOG_WARN("push_back failed", K(partition_stat), K(ret)); } } } if (OB_ITER_END == ret) { ret = OB_SUCCESS; } } // revert partition iterator if (NULL != partition_iter) { gctx_.par_ser_->revert_pg_iter(partition_iter); partition_iter = NULL; } } return ret; } int ObService::check_server_empty(const ObCheckServerEmptyArg& arg, const bool wait_log_scan, bool& is_empty) { int ret = OB_SUCCESS; is_empty = true; if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else { if (ObCheckServerEmptyArg::BOOTSTRAP == arg.mode_) { // "is_valid_heartbeat" is valid: // 1. RS is between "start_service" and "full_service", the server has not added to RS; // 2. RS is in "full_service" and the server has added to RS; // 3. To avoid misjudgment in scenario 1 while add server, this check is skipped here if (lease_state_mgr_.is_valid_heartbeat()) { LOG_WARN("server already in rootservice lease"); is_empty = false; } } // wait log scan finish if (is_empty && wait_log_scan) { const int64_t WAIT_LOG_SCAN_TIME_US = 2 * 1000 * 1000; // only wait 2s for empty server const int64_t SLEEP_INTERVAL_US = 500; const int64_t start_time_us = ObTimeUtility::current_time(); int64_t end_time_us = start_time_us; int64_t timeout_ts = THIS_WORKER.get_timeout_ts(); if (INT64_MAX == THIS_WORKER.get_timeout_ts()) { timeout_ts = start_time_us + WAIT_LOG_SCAN_TIME_US; } while (!stopped_ && !gctx_.par_ser_->is_scan_disk_finished()) { end_time_us = ObTimeUtility::current_time(); if (end_time_us > timeout_ts) { LOG_WARN("wait log scan finish timeout", K(timeout_ts), LITERAL_K(WAIT_LOG_SCAN_TIME_US)); is_empty = false; break; } usleep(static_cast(std::min(timeout_ts - end_time_us, SLEEP_INTERVAL_US))); } } if (is_empty) { if (!gctx_.par_ser_->is_empty()) { LOG_WARN("partition service is not empty"); is_empty = false; } } if (is_empty) { if (!OBSERVER.is_log_dir_empty()) { LOG_WARN("log dir is not empty"); is_empty = false; } } } return ret; } int ObService::report_replica() { int ret = OB_SUCCESS; LOG_INFO("receive report all replicas request"); if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else { ObIPartitionGroupIterator* partition_iter = NULL; ObIPartitionGroup* partition = NULL; int64_t replica_count = 0; if (NULL == (partition_iter = gctx_.par_ser_->alloc_pg_iter())) { ret = OB_ALLOCATE_MEMORY_FAILED; LOG_WARN("Fail to alloc partition iter, ", K(ret)); } else { while (OB_SUCC(ret)) { ObPartitionArray pkeys; if (OB_FAIL(partition_iter->get_next(partition))) { if (OB_ITER_END != ret) { LOG_WARN("Fail to get next partition, ", K(ret)); } } else if (NULL == partition) { ret = OB_ERR_UNEXPECTED; LOG_WARN("The partition is NULL, ", K(ret)); } else if (OB_FAIL(partition->get_all_pg_partition_keys(pkeys))) { LOG_WARN("get all pg partition keys error", "pg_key", partition->get_partition_key()); if (OB_ENTRY_NOT_EXIST != ret && OB_PARTITION_NOT_EXIST != ret) { LOG_WARN("get partition failed", K(ret)); } else { // The partition has been deleted. There is no need to trigger the report ret = OB_SUCCESS; } } else if (OB_FAIL(submit_pt_update_task( partition->get_partition_key(), true /*need report checksum*/, true /*with role report*/))) { if (OB_PARTITION_NOT_EXIST == ret) { // The GC thread is already working, // and deleted during traversal, the replica has been deleted needs to be avoided blocking the start process ret = OB_SUCCESS; LOG_INFO("this partition is already not exist", K(ret), "partition_key", partition->get_partition_key()); } else { LOG_WARN( "submit partition table update task failed", K(ret), "partition_key", partition->get_partition_key()); } } else { // Update partition meta table without concern for error codes submit_pg_pt_update_task(pkeys); ++replica_count; } } if (OB_ITER_END == ret) { ret = OB_SUCCESS; } } if (NULL != partition_iter) { gctx_.par_ser_->revert_pg_iter(partition_iter); } LOG_INFO("submit all replicas report", K(ret)); SERVER_EVENT_ADD("storage", "report_replica", K(ret), "replica_count", replica_count); } return ret; } int ObService::recycle_replica() { int ret = OB_SUCCESS; if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else { // gctx_.par_ser_->garbage_clean(); } return ret; } int ObService::clear_location_cache() { // TODO : implement after kvcache support clear interface int ret = OB_NOT_IMPLEMENT; return ret; } int ObService::drop_replica(const obrpc::ObDropReplicaArg& arg) { int ret = OB_SUCCESS; LOG_INFO("receive drop replica request"); if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else if (!arg.is_valid()) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid arg", K(arg), K(ret)); } else { ObRemoveNonPaxosReplicaArg rm_arg; rm_arg.key_ = arg.partition_key_; rm_arg.dst_ = arg.member_; rm_arg.switch_epoch_ = GCTX.get_switch_epoch2(); if (OB_FAIL(remove_non_paxos_replica(rm_arg))) { LOG_WARN("remove_replica failed", K(rm_arg), K(ret)); } } SERVER_EVENT_ADD("storage", "drop_replica", K(ret), "partition", arg.partition_key_); return ret; } int ObService::set_ds_action(const obrpc::ObDebugSyncActionArg& arg) { int ret = OB_SUCCESS; if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else if (!arg.is_valid()) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid arg", K(arg), K(ret)); } else if (OB_FAIL(GDS.set_global_action(arg.reset_, arg.clear_, arg.action_))) { LOG_WARN("set debug sync global action failed", K(ret), K(arg)); } return ret; } int ObService::update_cluster_info(ObClusterInfoArg& arg) { int ret = OB_SUCCESS; UNUSEDx(arg); return ret; } int ObService::request_heartbeat(ObLeaseRequest& lease_request) { int ret = OB_SUCCESS; if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else if (OB_FAIL(heartbeat_process_.init_lease_request(lease_request))) { LOG_WARN("init_lease_request failed", K(ret)); } return ret; } // only for bootstrap, Use new mode directly int ObService::broadcast_sys_schema(const ObSArray& table_schemas) { int ret = OB_SUCCESS; if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else if (table_schemas.count() <= 0) { ret = OB_INVALID_ARGUMENT; LOG_WARN("table_schemas is empty", K(table_schemas), K(ret)); } else if (OB_ISNULL(gctx_.schema_service_)) { ret = OB_ERR_UNEXPECTED; LOG_WARN("schema_service is null", K(ret)); } else if (OB_FAIL(gctx_.schema_service_->init_sys_schema(table_schemas))) { LOG_WARN("init_sys_schema failed", K(table_schemas), K(ret)); } else { } return ret; } // get tenant's refreshed schema version in new mode int ObService::get_tenant_refreshed_schema_version( const obrpc::ObGetTenantSchemaVersionArg& arg, obrpc::ObGetTenantSchemaVersionResult& result) { int ret = OB_SUCCESS; result.schema_version_ = OB_INVALID_VERSION; if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else if (!arg.is_valid()) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid arg", K(ret)); } else if (OB_ISNULL(gctx_.schema_service_)) { ret = OB_ERR_UNEXPECTED; LOG_WARN("schema_service is null", K(ret)); } else if (OB_FAIL(gctx_.schema_service_->get_tenant_refreshed_schema_version( arg.tenant_id_, result.schema_version_, false /*core_version*/, true /*use_new_mode*/))) { LOG_WARN("fail to get tenant refreshed schema version", K(ret), K(arg)); } return ret; } int ObService::check_partition_table() { int ret = OB_SUCCESS; if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else if (OB_FAIL(pt_checker_.schedule_task())) { LOG_WARN("pt_checker schedule_task failed", K(ret)); } return ret; } int ObService::sync_pg_partition_table(const obrpc::Int64& arg) { LOG_INFO("receive sync pg partition meta table request"); int ret = OB_SUCCESS; if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else if (arg <= 0) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid argument", K(ret), K(arg)); } else if (OB_FAIL(ObPGPartitionMTUpdater::get_instance().sync_pg_pt(arg))) { LOG_WARN("submit sync pg partition meta table failed", K(ret), K(arg)); } SERVER_EVENT_ADD("storage", "sync_pg_partition_table", K(ret), "version", arg); return ret; } int ObService::sync_partition_table(const obrpc::Int64& arg) { LOG_INFO("receive sync partition table request"); int ret = OB_SUCCESS; if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else if (arg <= 0) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid argument", K(ret), K(arg)); } else if (OB_FAIL(partition_table_updater_.sync_pt(arg))) { LOG_WARN("submit sync partition table failed", K(ret), K(arg)); } SERVER_EVENT_ADD("storage", "sync_partition_table", K(ret), "version", arg); return ret; } int ObService::check_dangling_replica_exist(const obrpc::Int64& arg) { LOG_INFO("receive check dangling replica request"); int ret = OB_SUCCESS; if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else if (arg <= 0) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid argument", K(ret), K(arg)); } else if (OB_FAIL(pt_checker_.schedule_task(arg))) { LOG_WARN("pt_checker schedule_task failed", K(ret), K(arg)); } SERVER_EVENT_ADD("storage", "check_dangling_replica_exist", K(ret), "version", arg); return ret; } int ObService::get_server_heartbeat_expire_time(int64_t& lease_expire_time) { int ret = OB_SUCCESS; if (OB_UNLIKELY(!inited_)) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else { lease_expire_time = lease_state_mgr_.get_heartbeat_expire_time(); } return ret; } bool ObService::is_heartbeat_expired() const { bool bret = false; // returns false on error if (OB_UNLIKELY(!inited_)) { LOG_WARN("not init"); } else { bret = !lease_state_mgr_.is_valid_heartbeat(); } return bret; } bool ObService::is_svr_lease_valid() const { // Determine if local lease is valid in OFS mode bool bret = false; if (OB_UNLIKELY(!inited_)) { LOG_WARN("not init"); } else { bret = lease_state_mgr_.is_valid_lease(); } return bret; } int64_t ObService::get_partition_table_updater_user_queue_size() const { return partition_table_updater_.get_partition_table_updater_user_queue_size(); } int64_t ObService::get_partition_table_updater_sys_queue_size() const { return partition_table_updater_.get_partition_table_updater_sys_queue_size(); } int64_t ObService::get_partition_table_updater_core_queue_size() const { return partition_table_updater_.get_partition_table_updater_core_queue_size(); } int ObService::set_tracepoint(const obrpc::ObAdminSetTPArg& arg) { int ret = OB_SUCCESS; if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else { if (arg.event_name_.length() > 0) { ObSqlString str; if (OB_FAIL(str.assign(arg.event_name_))) { LOG_WARN("string assign failed", K(ret)); } else { TP_SET_EVENT(str.ptr(), arg.error_code_, arg.occur_, arg.trigger_freq_); } } else { TP_SET_EVENT(arg.event_no_, arg.error_code_, arg.occur_, arg.trigger_freq_); } LOG_INFO("set event", K(arg)); } return ret; } int ObService::cancel_sys_task(const share::ObTaskId& task_id) { int ret = OB_SUCCESS; if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else if (task_id.is_invalid()) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid args", K(ret), K(task_id)); } else if (OB_FAIL(SYS_TASK_STATUS_MGR.cancel_task(task_id))) { LOG_WARN("failed to cancel sys task", K(ret), K(task_id)); } return ret; } int ObService::get_all_partition_status(int64_t& inactive_num, int64_t& total_num) const { int ret = OB_SUCCESS; if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else if (OB_FAIL(gctx_.par_ser_->get_all_partition_status(inactive_num, total_num))) { LOG_WARN("fail to get all partition status", K(ret)); } return ret; } int ObService::get_tenant_group_string(common::ObString& ttg_string) { int ret = OB_SUCCESS; ttg_string = GCONF.tenant_groups.str(); return ret; } int ObService::get_root_server_status(ObGetRootserverRoleResult& get_role_result) { int ret = OB_SUCCESS; // TODO : replace with LEADER && FOLLOWER. (same to ObRsMgr) const ObPartitionKey partition_key(combine_id(OB_SYS_TENANT_ID, OB_ALL_CORE_TABLE_TID), ObIPartitionTable::ALL_CORE_TABLE_PARTITION_ID, ObIPartitionTable::ALL_CORE_TABLE_PARTITION_NUM); ObPartitionReplica replica; get_role_result.replica_.reset(); if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else if (OB_ISNULL(gctx_.root_service_)) { ret = OB_ERR_UNEXPECTED; LOG_WARN("get invalid root_service", K(ret)); } else if (OB_FAIL(fill_partition_replica(partition_key, replica))) { if (OB_PARTITION_NOT_EXIST == ret) { ret = OB_SUCCESS; get_role_result.type_ = REPLICA_TYPE_MAX; get_role_result.role_ = ObRoleMgr::SLAVE; get_role_result.zone_ = GCONF.zone.str(); get_role_result.status_ = status::INVALID; } else { LOG_WARN("fail to fill partition replica", K(ret), K(partition_key)); } } else if (OB_FAIL(get_role_result.replica_.assign(replica))) { LOG_WARN("fail to assign replica", K(ret), K(replica)); } else { get_role_result.type_ = replica.replica_type_; get_role_result.role_ = gctx_.root_service_->in_service() ? ObRoleMgr::OB_MASTER : ObRoleMgr::SLAVE; get_role_result.zone_ = GCONF.zone.str(); get_role_result.status_ = gctx_.root_service_->get_status(); if (gctx_.root_service_->in_service()) { if (OB_FAIL(gctx_.root_service_->get_root_partition(get_role_result.partition_info_))) { LOG_WARN("fail to get root partition", KR(ret)); } } } return ret; } int ObService::get_master_root_server(obrpc::ObGetRootserverRoleResult& result) { int ret = OB_SUCCESS; ObAddr master_rs; bool found = false; if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("not init", K(ret)); } else if (OB_ISNULL(gctx_.rs_mgr_)) { ret = OB_ERR_UNEXPECTED; LOG_WARN("rs mgr is null", K(ret)); } else if (OB_NOT_NULL(gctx_.root_service_)) { // If itself is RS, it is no need to get RS by RPC if (gctx_.root_service_->in_service()) { if (OB_FAIL(get_root_server_status(result))) { LOG_WARN("failed to get rootserver status", K(ret), K(master_rs)); } else { found = true; } } } if (OB_FAIL(ret)) { } else if (found) { // If itself is RS, it is no need to get RS by RPC } else if (OB_FAIL(gctx_.rs_mgr_->get_master_root_server(master_rs))) { LOG_WARN("failed to find master rs", K(ret), K(master_rs)); } else if (gctx_.self_addr_ == master_rs) { // The master_rs stored in rs_mgr must not be local, // it has been judged that the itself is not master_rs ret = OB_RS_NOT_MASTER; LOG_WARN("not master rootserver", K(ret), K(master_rs)); } else if (OB_ISNULL(gctx_.rs_rpc_proxy_)) { ret = OB_INVALID_ARGUMENT; LOG_WARN("common rpc proxy is null", K(ret)); } else if (OB_FAIL(gctx_.rs_rpc_proxy_->to_addr(master_rs).get_root_server_status(result))) { LOG_WARN("failed to get rootserver role", K(ret), K(master_rs)); } return ret; } int ObService::refresh_core_partition() { int ret = OB_SUCCESS; const int64_t table_id = combine_id(OB_SYS_TENANT_ID, OB_ALL_CORE_TABLE_TID); const int64_t partition_id = ObIPartitionTable::ALL_CORE_TABLE_PARTITION_ID; ObPartitionLocation location; bool is_cache_hit = false; int64_t expire_renew_time = INT64_MAX; if (OB_FAIL(GCTX.location_cache_->get(table_id, partition_id, location, expire_renew_time, is_cache_hit))) { LOG_WARN("fail to refresh core partition", K(ret)); } else { #if !defined(NDEBUG) LOG_INFO("refresh core partition success", K(location)); #endif } return ret; } int ObService::split_partition(const obrpc::ObSplitPartitionArg& split_info, obrpc::ObSplitPartitionResult& result) { int ret = OB_SUCCESS; result.reset(); if (!split_info.is_valid()) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid argument", K(ret), K(split_info)); } else if (OB_FAIL(gctx_.par_ser_->split_partition(split_info.split_info_, result.get_result()))) { LOG_WARN("fail to split partition", K(ret)); } return ret; } int ObService::stop_partition_write(const obrpc::Int64& switchover_timestamp, obrpc::Int64& result) { // TODO for switchover int ret = OB_SUCCESS; result = switchover_timestamp; return ret; } int ObService::check_partition_log(const obrpc::Int64& switchover_timestamp, obrpc::Int64& result) { // Check that the log of all replicas in local have reached synchronization status // The primary has stopped writing int ret = OB_SUCCESS; int64_t balance_task = ObDagScheduler::get_instance().get_dag_count(ObIDag::DAG_TYPE_GROUP_MIGRATE); if (balance_task > 0) { ret = OB_EAGAIN; result = switchover_timestamp; LOG_INFO("observer already has task to do", K(switchover_timestamp), K(balance_task)); } else if (OB_FAIL(gctx_.par_ser_->check_all_partition_sync_state(switchover_timestamp))) { LOG_WARN("fail to check_all_partition_sync_state", K(ret)); } else { result = switchover_timestamp; } return ret; } int ObService::batch_set_member_list(const obrpc::ObBatchStartElectionArg& arg, obrpc::Int64& result) { UNUSED(arg); UNUSED(result); int ret = OB_NOT_SUPPORTED; return ret; } int ObService::batch_write_cutdata_clog(const obrpc::ObBatchWriteCutdataClogArg& arg, obrpc::ObBatchCheckRes& result) { UNUSED(result); LOG_INFO("receive batch write cutdata clog request", "task_count", arg.pkeys_.count(), "epoch", arg.index_); #if !defined(NDEBUG) for (int64_t i = 0; i < arg.pkeys_.count(); i++) { LOG_INFO("output pkey", K(i), K(arg.pkeys_.at(i))); } #endif int ret = OB_NOT_SUPPORTED; return ret; } int ObService::batch_wait_leader(const obrpc::ObBatchCheckLeaderArg& arg, obrpc::ObBatchCheckRes& result) { int ret = OB_SUCCESS; int64_t start_time = ObTimeUtility::current_time(); LOG_INFO("receive batch wait leader request", "task_count", arg.pkeys_.count(), "epoch", arg.index_); #if !defined(NDEBUG) for (int64_t i = 0; i < arg.pkeys_.count(); i++) { LOG_INFO("output pkey", K(i), K(arg.pkeys_.at(i))); } #endif if (!arg.is_valid()) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid argument", K(ret), K(arg)); } else if (OB_FAIL(gctx_.par_ser_->batch_check_leader_active(arg, result))) { LOG_WARN("fail to batch_check_leader_active", K(ret)); } else { // do nothing } result.index_ = arg.index_; int64_t cost = ObTimeUtility::current_time() - start_time; LOG_INFO("batch_wait_leader finished", KR(ret), K(cost), "task_count", arg.pkeys_.count()); return ret; } int ObService::estimate_partition_rows(const obrpc::ObEstPartArg& arg, obrpc::ObEstPartRes& res) const { int ret = OB_SUCCESS; LOG_DEBUG("receive estimate rows request", K(arg)); if (!inited_) { ret = OB_INVALID_ARGUMENT; LOG_WARN("service is not inited", K(ret)); } else if (OB_FAIL( ObOptEstCost::storage_estimate_rowcount(gctx_.par_ser_, ObStatManager::get_instance(), arg, res))) { LOG_WARN("failed to estimate partition rowcount", K(ret)); } return ret; } int ObService::get_wrs_info(const obrpc::ObGetWRSArg& arg, obrpc::ObGetWRSResult& result) { int ret = OB_SUCCESS; const int64_t tenant_id = arg.tenant_id_; const bool need_filter = arg.need_filter_; const bool need_inner_table = (ObGetWRSArg::USER_TABLE != arg.scope_); const bool need_user_table = (ObGetWRSArg::INNER_TABLE != arg.scope_); result.reset(); if (OB_FAIL(get_tenant_replica_wrs_info( result.replica_wrs_info_list_, tenant_id, need_filter, need_inner_table, need_user_table))) { LOG_WARN("fail to get tenant replica wrs info", KR(ret), K(tenant_id), K(need_filter), K(need_inner_table), K(need_user_table)); } result.err_code_ = ret; result.self_addr_ = gctx_.self_addr_; return OB_SUCCESS; } int ObService::check_physical_flashback_succ( const obrpc::ObCheckPhysicalFlashbackArg& arg, obrpc::ObPhysicalFlashbackResultArg& result) { int ret = OB_SUCCESS; if (!inited_) { ret = OB_INVALID_ARGUMENT; LOG_WARN("service is not inited", K(ret)); } else if (OB_FAIL(gctx_.par_ser_->check_physical_flashback_succ(arg, result))) { LOG_WARN("fail to check physical flashback", K(ret), K(arg)); } return ret; } int ObService::update_pg_backup_task_info(const common::ObIArray& pg_task_info_array) { int ret = OB_SUCCESS; ObTenantBackupTaskInfo task_info; ObTenantBackupTaskInfo dest_task_info; ObMySQLTransaction trans; const int64_t MAX_EXECUTE_TIMEOUT_US = 5 * 60L * 1000 * 1000; // 5min int64_t stmt_timeout = MAX_EXECUTE_TIMEOUT_US; ObTimeoutCtx timeout_ctx; int64_t macro_block_count = 0; int64_t finish_partition_count = 0; int64_t finish_macro_block_count = 0; int64_t input_bytes = 0; int64_t output_bytes = 0; if (!inited_) { ret = OB_NOT_INIT; LOG_WARN("service do not init", K(ret)); } else if (pg_task_info_array.empty()) { ret = OB_INVALID_ARGUMENT; LOG_WARN("pg backup status is invalid", K(ret), K(pg_task_info_array)); } else { for (int64_t i = 0; i < pg_task_info_array.count(); ++i) { const ObPGBackupTaskInfo& pg_task_info = pg_task_info_array.at(i); if (OB_SUCCESS != pg_task_info.result_) { // do nothing } else { finish_partition_count += pg_task_info.finish_partition_count_; macro_block_count += pg_task_info.macro_block_count_; finish_macro_block_count += pg_task_info.finish_macro_block_count_; input_bytes += pg_task_info.input_bytes_; output_bytes += pg_task_info.output_bytes_; } } if (OB_FAIL(timeout_ctx.set_trx_timeout_us(stmt_timeout))) { LOG_WARN("fail to set trx timeout", K(ret), K(stmt_timeout)); } else if (OB_FAIL(timeout_ctx.set_timeout(stmt_timeout))) { LOG_WARN("set timeout context failed", K(ret)); } else if (OB_FAIL(trans.start(GCTX.sql_proxy_))) { LOG_WARN("failed to start trans", K(ret)); } else { const uint64_t tenant_id = extract_tenant_id(pg_task_info_array.at(0).table_id_); if (OB_FAIL(tenant_backup_task_updater_.get_tenant_backup_task(tenant_id, trans, task_info))) { LOG_WARN("failed to get tenant backup task info", K(ret), K(tenant_id)); } else if (ObTenantBackupTaskInfo::DOING != task_info.status_ && ObTenantBackupTaskInfo::CANCEL != task_info.status_) { ret = OB_ERR_UNEXPECTED; LOG_WARN("tenant backup task status is unexpected", K(ret), K(task_info)); } else if (OB_FAIL(pg_backup_task_updater_.update_status_and_result_and_statics(pg_task_info_array))) { LOG_WARN("failed to update pg backup task status", K(ret), K(pg_task_info_array)); } else { dest_task_info = task_info; dest_task_info.finish_partition_count_ += finish_macro_block_count; dest_task_info.macro_block_count_ += macro_block_count; dest_task_info.finish_macro_block_count_ += finish_macro_block_count; dest_task_info.input_bytes_ += input_bytes; dest_task_info.output_bytes_ += output_bytes; if (OB_FAIL(tenant_backup_task_updater_.update_tenant_backup_task(trans, task_info, dest_task_info))) { LOG_WARN("failed to update tenant backup task", K(ret), K(task_info), K(dest_task_info)); } } if (OB_SUCC(ret)) { if (OB_FAIL(trans.end(true /*commit*/))) { OB_LOG(WARN, "failed to commit", K(ret)); } } else { int tmp_ret = OB_SUCCESS; if (OB_SUCCESS != (tmp_ret = trans.end(false /*rollback*/))) { OB_LOG(WARN, "failed to rollback", K(ret), K(tmp_ret)); } } } } return ret; } int ObService::refresh_memory_stat() { return ObDumpTaskGenerator::generate_mod_stat_task(); } int ObService::renew_in_zone_hb(const share::ObInZoneHbRequest& arg, share::ObInZoneHbResponse& result) { int ret = OB_SUCCESS; UNUSED(arg); UNUSED(result); return ret; } int ObService::pre_process_server_reply(const obrpc::ObPreProcessServerReplyArg& arg) { int ret = OB_SUCCESS; UNUSED(arg); return ret; } int ObService::broadcast_rs_list(const ObRsListArg& arg) { int ret = OB_SUCCESS; if (!arg.is_valid()) { ret = OB_INVALID_ARGUMENT; LOG_WARN("invalid argument", KR(ret), K(arg)); } else if (OB_UNLIKELY(!inited_)) { ret = OB_NOT_INIT; LOG_WARN("service do not init", KR(ret), K(arg)); } else if (OB_ISNULL(GCTX.rs_mgr_)) { ret = OB_ERR_UNEXPECTED; LOG_WARN("rs mgr is null", KR(ret), K(arg)); } else if (OB_FAIL(GCTX.rs_mgr_->force_set_master_rs(arg.master_rs_))) { LOG_WARN("fail to set master rs", KR(ret), K(arg)); } else { LOG_INFO("observer set master rs success", K(arg)); } return ret; } } // end namespace observer } // end namespace oceanbase