Oceanbase数据库中同Oracle一样存在一些以“_”开头的隐藏参数，用于功能的开关和微调等作用。目前v4.2.5.3 有491个参数，分为集群级和租户级。隐藏参数226个。通过数据库参数可以研究该版本的功能特性。
obclient(root@sys)[oceanbase]> select version();
+---------------------------+
| version() |
+---------------------------+
| 5.7.25-OceanBase-v4.2.5.3 |
+---------------------------+
1 row in set (0.003 sec)
obclient(root@sys)[oceanbase]> select scope,count(*) from (select distinct name,scope from __all_virtual_sys_parameter_stat ) group by scope;
+---------+----------+
| scope | count(*) |
+---------+----------+
| CLUSTER | 253 |
| TENANT | 238 |
+---------+----------+
obclient(root@sys)[oceanbase]> select distinct name,scope,info, value from __all_virtual_sys_parameter_stat where regexp_like(name,'^_.*');
+-------------------------------------------------+---------+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+------------------------------------------------------------------------------------+
| name | scope | info | value |
+-------------------------------------------------+---------+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+------------------------------------------------------------------------------------+
| _dtl_tenant_mem_limit | CLUSTER | the buffer size of storage stream rpcRange: [1B,) | 1G |
| _enable_auth_switch | CLUSTER | Control whether to use auth_switch. | True |
| _enable_unit_gc_wait | CLUSTER | Used to control enable or disable the unit smooth gc feature, enabled by default. | True |
| _ob_pl_compile_max_concurrency | CLUSTER | The maximum number of threads that an observer node can compile PL concurrently. | 4 |
| _enable_dbms_job_package | CLUSTER | Control whether can use DBMS_JOB package. | False |
| _enable_memleak_light_backtrace | CLUSTER | specifies whether allow memleak to get the backtrace of malloc by light_backtrace | True |
| _faststack_min_interval | CLUSTER | Minimum interval for OBServer to automatically collect the obstack. Default: 30min. Range: [1s,+∞) | 30m |
| _faststack_req_queue_size_threshold | CLUSTER | When the size of the req_queue reaches this threshold, the obstack will be collected automatically. Default: 0, means set off. Range: [0, +∞) | 0 |
| _ha_diagnose_history_recycle_interval | CLUSTER | The recycle interval time of diagnostic history data. Range: [2m, 180d] | 7d |
| _obkv_feature_mode | CLUSTER | _obkv_feature_mode is a option list to control specified OBKV features on/off. | |
| _enable_persistent_compiled_routine | CLUSTER | specifies whether the feature of storeing dll to disk is turned on. The default value is TRUE. Value: TRUE: turned on FALSE: turned off | True |
| _enable_di_experimental_feature_flags | CLUSTER | [0, +∞)enable diagnostic info experimental feature | 0 |
| _enable_diagnostic_info_cache | CLUSTER | enable diagnostic info cache | True |
| _ob_ash_disk_write_enable | CLUSTER | enable active session history early flush | True |
| _ob_ash_enable | CLUSTER | enable active session history | True |
| _ob_ash_size | CLUSTER | to limit the memory size for ash buffer. Range: [0,1G] 0 means using default ash size, 30MB in normal case, 10M in mini mode | 0M |
| _force_malloc_for_absent_tenant | CLUSTER | force malloc even if tenant does not exist in observer | False |
| _schema_memory_recycle_interval | CLUSTER | the time interval between the schedules of schema memory recycle task. 0 means only turn off gc current allocator, and other schema memory recycle task's interval will be 15mins. Range [0s,) | 15m |
| _force_explict_500_malloc | CLUSTER | Force 500 memory for explicit allocation | False |
| _enable_backtrace_function | CLUSTER | Decide whether to let the backtrace function take effect | True |
| _enable_reserved_user_dcl_restriction | CLUSTER | specifies whether to forbid non-reserved user to modify reserved users | False |
| _enable_px_fast_reclaim | CLUSTER | Enable the fast reclaim function through PX tasks deteting for survival by detect manager. The default value is True. | True |
| _load_tde_encrypt_engine | CLUSTER | load the engine that meet the security classification requirement to encrypt data. default NONE | NONE |
| _enable_protocol_diagnose | CLUSTER | enables protocol layer diagnosis. The default value is False. | True |
| _display_mysql_version | CLUSTER | dynamic mysql version of mysql mode observer | 5.7.25 |
| _ob_enable_direct_load | CLUSTER | Enable or disable direct path load | True |
| _ob_plan_cache_auto_flush_interval | CLUSTER | time interval for auto periodic flush plan cache. Range: [0s, +∞) | 0s |
| _advance_checkpoint_timeout | CLUSTER | the timeout for backup/migrate advance checkpoint Range: [10s,180m] | 30m |
| _endpoint_tenant_mapping | CLUSTER | This parameter will store the mapping of endpoint and tenant | |
| _enable_tenant_sql_net_thread | CLUSTER | Dispatch mysql request to each tenant with True, or disable with False | True |
| _ignore_system_memory_over_limit_error | CLUSTER | When the hold of observer tenant is over the system_memory, print ERROR with False, or WARN with True | False |
| _enable_new_sql_nio | CLUSTER | specifies whether SQL serial network is turned on. Turned on to support mysql_send_long_dataThe default value is FALSE. Value: TRUE: turned on FALSE: turned off | True |
| _session_context_size | CLUSTER | limits the total number of (namespace, attribute) pairs used by all application contexts in the user session. | 10000 |
| _enable_newsort | CLUSTER | control if enable encode sort | True |
| _ob_obj_dep_maint_task_interval | CLUSTER | The execution interval of the task of maintaining the dependency of the object. Range: [0, 10s] | 1ms |
| _ob_enable_fast_parser | CLUSTER | control if enable fast parser | True |
| _enable_trace_session_leak | CLUSTER | specifies whether to enable tracing session leak | False |
| _enable_block_file_punch_hole | CLUSTER | specifies whether to punch whole when free blocks in block_file | False |
| _resource_limit_spec | CLUSTER | this parameter encodes some resource limit parameters to json | auto |
| _enable_resource_limit_spec | CLUSTER | specify whether resource limit check is turned on | False |
| _enable_px_ordered_coord | CLUSTER | enable px task ordered coord | False |
| _send_bloom_filter_size | CLUSTER | Set send bloom filter slice sizeRange: [1, +∞) | 1024 |
| _enable_easy_keepalive | CLUSTER | enable keepalive for each TCP connection. | True |
| _xa_gc_interval | CLUSTER | specifies the scan interval of the gc worker | 1h |
| _xa_gc_timeout | CLUSTER | specifies the threshold value for a xa record to be considered as obsolete | 24h |
| _ob_ssl_invited_nodes | CLUSTER | when rpc need use ssl, we will use it to store invited server ipv4 during grayscale change.when it is finish, it can use ALL instead of all server ipv4 | NONE |
| _enable_oracle_priv_check | CLUSTER | whether turn on oracle privilege check | True |
| _enable_plan_cache_mem_diagnosis | CLUSTER | wether turn plan cache ref count diagnosis on | False |
| _ob_plan_cache_gc_strategy | CLUSTER | OFF: disabled, REPORT: check leaked cache object infos only, AUTO: check and release leaked cache obj. | REPORT |
| _upgrade_stage | CLUSTER | specifies the upgrade stage. NONE: in non upgrade stage, PREUPGRADE: in pre upgrade stage, DBUPGRADE: in db uprade stage, POSTUPGRADE: in post upgrade stage. | NONE |
| _ob_enable_prepared_statement | CLUSTER | control if enable prepared statement | True |
| _backup_idle_time | CLUSTER | the time interval between the schedules of physical backup task. Range: [10s, +∞) | 5m |
| _restore_idle_time | CLUSTER | the time interval between the schedules of physical restore task. Range: [10s, +∞) | 1m |
| _bloom_filter_ratio | CLUSTER | the px bloom filter false-positive rate.the default value is 1, range: [0,100] | 35 |
| _ob_ddl_timeout | CLUSTER | the config parameter of ddl timeoutRange: [1s, +∞) | 1000s |
| _px_max_pipeline_depth | CLUSTER | max parallel execution pipeline depth, range: [2,3] | 2 |
| __easy_memory_reserved_percentage | CLUSTER | the percentage of easy memory reserved size. The default value is 0. Range: [0,100] | 0 |
| __easy_memory_limit | CLUSTER | max memory size which can be used by libeasy. The default value is 4G. Range: [256M,) | 4G |
| _max_elr_dependent_trx_count | CLUSTER | max elr dependent transaction count | 0 |
| _cache_wash_interval | CLUSTER | specify interval of cache background wash | 200ms |
| _migrate_block_verify_level | CLUSTER | specify what kind of verification should be done when migrating macro block. 0 : no verification will be done 1 : physical verification2 : logical verification | 1 |
| _ob_elr_fast_freeze_threshold | CLUSTER | per row update counts threshold to trigger minor freeze for tables with ELR optimization | 500000 |
| _fast_commit_callback_count | CLUSTER | [0,)trigger max callback count allowed within transaction for durable callback checkpoint, 0 represents not allow durable callbackRange: [0, not limited callback count | 10000 |
| _private_buffer_size | CLUSTER | [0B,)the trigger remaining data size within transaction for immediate logging, 0B represents not trigger immediate loggingRange: [0B, total size of memory] | 16K |
| _force_skip_encoding_partition_id | CLUSTER | force the specified partition to major without encoding row store, only for emergency! | |
| _enable_compaction_diagnose | CLUSTER | enable compaction diagnose functionValue: True:turned on; False: turned off | False |
| _data_storage_io_timeout | CLUSTER | io timeout for data storage, Range [1s,600s]. The default value is 10s | 10s |
| _auto_broadcast_tablet_location_rate_limit | CLUSTER | Maximum number of tablets broadcasted per second by a single observer. When the value is 0, it means shutting down related logic. | 10000 |
| _auto_refresh_tablet_location_interval | CLUSTER | Polling period of auto refresh tablet location service. When the value is 0, it means shutting down related service. Range: [0s, +∞) | 10m |
| _server_standby_fetch_log_bandwidth_limit | CLUSTER | the max bandwidth in bytes per second that can be occupied by the synchronizing log from primary cluster of a server in standby cluster | 0MB |
| _max_rpc_packet_size | CLUSTER | the max rpc packet size when sending RPC or responding RPC results | 2047MB |
| _enable_pkt_nio | CLUSTER | enable pkt-nio, the new RPC frameworkValue: True:turned on; False: turned off | True |
| _ob_get_gts_ahead_interval | CLUSTER | get gts ahead interval. Range: [0s, 1s] | 0s |
| _parallel_redo_logging_trigger | CLUSTER | size of single transaction's pending redo log to trigger parallel writes redo log. Range: [0B,+∞) | 16M |
| _enable_parallel_redo_logging | CLUSTER | enable parallel write redo log. | True |
| _ob_trans_rpc_timeout | CLUSTER | transaction rpc timeout(s). Range: [0s, 3600s] | 3s |
| _rpc_checksum | CLUSTER | Force: always verify; Optional: verify when rpc_checksum non-zero; Disable: ignore verify | Force |
| _lcl_op_interval | CLUSTER | Scan interval for every detector node, smaller interval support larger deadlock scale, but cost more system resource. 0ms means disable deadlock, default value is 30ms. Range:[0ms, 1s] | 30ms |
| __min_full_resource_pool_memory | CLUSTER | the min memory value which is specified for a full resource pool. | 5368709120 |
| __balance_controller | CLUSTER | specifies whether the balance events are turned on or turned off. | |
| _ob_flush_log_at_trx_commit | CLUSTER | specifies whether to use logstore. The default value is 1.0 : flush logs asynchronously by logstore1 : flush logs to disk directly and synchronouslythis value can be set 0 only when _ob_logstore_service_addr is valid | 1 |
| _ob_logstore_service_addr | CLUSTER | the logstore server addr, which is expected to be valid when _ob_flush_log_at_trx_commit is 0 | |
| _recyclebin_object_purge_frequency | CLUSTER | the time to purge recyclebin. Range: [0m, +∞) | 10m |
| _px_chunklist_count_ratio | CLUSTER | the ratio of the dtl buffer manager list. Range: [1, 128] | 1 |
| _ob_max_thread_num | CLUSTER | ob max thread number upper limit of observer thread count. Range: [0, 10000), 0 means no limit. | 0 |
| _hidden_sys_tenant_memory | CLUSTER | the size of the memory reserved for hidden sys tenant, 0M means follow the adjusting value. | 0M |
| _enable_values_table_folding | CLUSTER | whether enable values statement folds self params | True |
| _max_malloc_sample_interval | CLUSTER | the max malloc times between two samples, which is not less than _min_malloc_sample_interval. 1 means to sample all malloc, Range: [1, 10000] | 256 |
| _min_malloc_sample_interval | CLUSTER | the min malloc times between two samples, which is not more than _max_malloc_sample_interval. 10000 means not to sample any malloc, Range: [1, 10000] | 16 |
| _auto_drop_recovering_auxiliary_tenant | CLUSTER | control whether to delete auxiliary tenant after recovering tables failed | True |
| _sql_insert_multi_values_split_opt | CLUSTER | True means that the split + batch optimization for inserting multiple rows of the insert values statement can be done | True |
| _delay_resource_recycle_after_correctness_issue | CLUSTER | whether hinder the recycling of the log resources and the sstable resources under correctness issues | False |
| _enable_defensive_check | CLUSTER | specifies whether allow to do some defensive checks when the query is executed, 0 means defensive check is disabled, 1 means normal defensive check is enabled, 2 means more strict defensive check is enabled, such as check partition id validity | 1 |
| _enable_partition_level_retry | CLUSTER | specifies whether allow the partition level retry when the leader changes | True |
| _chunk_row_store_mem_limit | CLUSTER | the maximum size of memory used by ChunkRowStore, 0 means follow operator's setting. Range: [0, +∞) | 0M |
| _datafile_usage_lower_bound_percentage | CLUSTER | the percentage of disk space usage lower bound to trigger datafile shrink. Range: [5,99] in integer | 10 |
| _datafile_usage_upper_bound_percentage | CLUSTER | the percentage of disk space usage upper bound to trigger datafile extend. Range: [5,99] in integer | 90 |
| _storage_stream_rpc_buffer_size | TENANT | [2M,128M]the buffer size of storage stream rpcRange: [2M, 128M] | 2M |
| _enable_sql_audit_query_sql | TENANT | Control whether to use sql_audit_query_sql. | False |
| _system_trig_enabled | TENANT | Enable or disable system trigger feature. | True |
| _enable_drop_and_add_index | TENANT | it specifies that whether we can drop and add index in single statement | False |
| _disable_record_outrow_lob_in_clog | TENANT | Whether to force outrow lob data to be written to CLOG for OBCDC synchronization when lob columns are not updated | False |
| _enable_check_trigger_const_variables_assign | TENANT | Used to control whether an error is reported when assigning a value to a const variable in a trigger under an Oracle tenant | True |
| _enable_ddl_worker_isolation | TENANT | a switch controling ddl thread isolation | False |
| _enable_dbms_lob_partial_update | TENANT | Enable the capability of dbms_lob to perform partial updates on LOB | False |
| _allow_skip_replay_redo_after_detete_tablet | TENANT | allow skip replay invalid redo log after tablet delete transaction is committed.The default value is FALSE. Value: TRUE means we allow skip replaying this invalid redo log, False means we do not alow such behavior. | False |
| _enable_choose_migration_source_policy | TENANT | Control whether to use chose_migration_source_policy. If the value of configure is false, it will not use chose_migration_source_policy and choose replica with the largest checkpoint scn as the source. | True |
| _query_record_size_limit | TENANT | set sql_audit and plan stat query sql size. Range: [0,67108864] in integer in integer. | 65536 |
| _multimodel_memory_trace_level | TENANT | Multi-mode memory tracking mechanism | 0 |
| _checkpoint_diagnose_preservation_count | TENANT | the count of checkpoint diagnose info preservation | 100 |
| _preserve_order_for_pagination | TENANT | enable preserver order for limit | False |
| _enable_new_query_range_extraction | TENANT | decide whether use new algorithm to extract query range. | True |
| _enable_range_extraction_for_not_in | TENANT | Enable extract query range for not in predicate | True |
| _iut_stat_collection_type | TENANT | specify index table usage stat collection type, values: SAMPLED, ALL | SAMPLED |
| _iut_max_entries | TENANT | maximum of index entries to be monitoring. | 30000 |
| _iut_enable | TENANT | specifies whether allow the index table usage start monitoring. | True |
| _standby_max_replay_gap_time | TENANT | The difference in replayable_scn between log streams on standby tenants is not greater than _standby_max_replay_gap_time, and the gap between sync_scn and replayable_scn of each log stream is kept reasonably small. Range: [10s, ) | 900s |
| _enable_inner_session_mgr | TENANT | enable/disable inner session mgr | True |
| _ob_sqlstat_enable | TENANT | enable/disable sql stat | True |
| _inlist_rewrite_threshold | TENANT | [1, 2147483647]specifies transform how much const params in IN list to values table | 1000 |
| _ha_rpc_timeout | TENANT | the rpc timeout for storage high availability. Range:[0, 120s] | 0 |
| _ha_tablet_info_batch_count | TENANT | the number of tablet replica info sent by on rpc for ha. Range: [0, +∞) in integer | 0 |
| _ls_migration_wait_completing_timeout | TENANT | the wait timeout in ls complete migration phase | 30m |
| _optimizer_skip_scan_enabled | TENANT | enable/disable index skip scan | False |
| _optimizer_better_inlist_costing | TENANT | enable improved costing of index access using in-list(s) | True |
| _stall_threshold_for_dynamic_worker | TENANT | threshold of dynamic worker works | 3ms |
| _enable_in_range_optimization | TENANT | Enable extract query range optimization for in predicate | True |
| _rebuild_replica_log_lag_threshold | TENANT | size of clog files that a replica lag behind leader to trigger rebuild, 0 means never trigger rebuild on purpose. Range: [0, +∞) | 0M |
| _wait_interval_after_parallel_ddl | TENANT | time interval for waiting other servers to refresh schema after parallel ddl is done | 30s |
| _complex_cbqt_table_num | TENANT | cost-based transform will be disabled when table count in a single stmt exceeds threshold | 10 |
| _optimizer_group_by_placement | TENANT | enable group by placement transform rule | True |
| _xsolapi_generate_with_clause | TENANT | OLAP API generates WITH clause | True |
| _max_dblink_conn_per_observer | TENANT | The maximum limit on the number of connections that can be opened simultaneously for a specific observer for any DBLink, default value is 256 | 256 |
| _with_subquery | TENANT | WITH subquery transformation,0: optimizer,1: materialize,2: inline | 0 |
| _enable_dblink_reuse_connection | TENANT | specifies whether dblink reuse connection in a session | True |
| _enable_hgby_llc_ndv_adaptive | TENANT | specifies whether llc ndv adptive is activated | True |
| _enable_hgby_skew_detection | TENANT | specifies whether hgby skew detection is enabled | True |
| _balance_wait_killing_transaction_end_threshold | TENANT | the threshold for waiting time after killing transactions until they end.Range: [10ms, 60s] | 100ms |
| _balance_kill_transaction_threshold | TENANT | the time given to the transaction to execute when do balancebefore it will be killed. Range: [1ms, 60s] | 100ms |
| _enable_balance_kill_transaction | TENANT | Specifies whether balance should actively kill transaction | False |
| _pipelined_table_function_memory_limit | TENANT | pipeline table function result set memory size limit. default 524288000 (500M), Range: [1024,18446744073709551615] | 524288000 |
| _enable_wait_remote_lock | TENANT | enable remote execution wait in lock wait mgr when lock conflict occurs | True |
| _enable_transaction_internal_routing | TENANT | enable SQLs of transaction routed to any servers in the cluster on demand | True |
| _px_join_skew_minfreq | TENANT | sets minimum frequency(%) for skewed value for parallel joins. Range: [1, 100] in integer | 30 |
| _px_join_skew_handling | TENANT | enables skew handling for parallel joins. The default value is True. | False |
| _transfer_task_retry_interval | TENANT | Retry interval after transfer task failure. Range: [0s, +∞). Default: 1m | 1m |
| _enable_active_txn_transfer | TENANT | Specifies whether support transfer active tx | False |
| _transfer_task_tablet_count_threshold | TENANT | Threshold for the count of tablets that can be processed by a transfer task. Range: (0, +∞) | 100 |
| _transfer_process_lock_tx_timeout | TENANT | transaction timeout for locking and unlocking transfer task. Range: [30s, +∞) | 100s |
| _transfer_service_wakeup_interval | TENANT | transfer service wakeup interval in errsim mode. Range: [1s, 5m] | 5m |
| _transfer_start_retry_count | TENANT | the number of transfer start retry. Range: [0, 64] in integer | 3 |
| _transfer_start_trans_timeout | TENANT | transfer start transaction timeout, Range [1ms,600s]. The default value is 1s | 1s |
| _transfer_finish_trans_timeout | TENANT | transfer finish transaction timeout, Range [1s,600s]. The default value is 10s | 10s |
| _transfer_start_rpc_timeout | TENANT | transfer start status rpc check some status ready timeout, Range [1ms,600s]. The default value is 10s | 10s |
| _enable_var_assign_use_das | TENANT | enable use das if the sql statement has variable assignment | False |
| _enable_insertup_replace_gts_opt | TENANT | By default, insert/replace ... values statement can use gts optimization | True |
| _enable_parallel_das_dml | TENANT | By default, the das service is allowed to use multiple threads to submit das tasks | False |
| _print_sample_ppm | TENANT | In the full link diagnosis, control the frequency of printing traces to the log (unit is ppm, parts per million). | 0 |
| _trace_control_info | TENANT | persistent control information for full-link trace | {"type_t":{"level":1,"sample_pct":"0.10","record_policy":"SAMPLE_AND_SLOW_QUERY"}} |
| _follower_snapshot_read_retry_duration | TENANT | the waiting time after the first judgment failure of strong reading on followerRange: [0ms, +∞) | 0ms |
| _px_object_sampling | TENANT | [1, 100000]parallel query sampling for base objects (100000 = 100%) | 200 |
| _px_bloom_filter_group_size | TENANT | specifies the px bloom filter each group size in sending to the other sqcRange: [1, +∞) or auto, the default value is auto | auto |
| _resource_limit_max_session_num | TENANT | the maximum number of sessions that can be created concurrently | 0 |
| _parallel_max_active_sessions | TENANT | max active parallel sessions allowed for tenant. Range: [0,+∞) | 0 |
| _enable_das_batch_rescan_flag | TENANT | enable das batch rescan for multiple scenarios. | 0 |
| _enable_distributed_das_scan | TENANT | enable distributed DAS scan | True |
| _enable_spf_batch_rescan | TENANT | enable das batch rescan for subplan filter | False |
| _enable_px_batch_rescan | TENANT | enable px batch rescan for nlj or subplan filter | True |
| _enable_enhanced_cursor_validation | TENANT | enable enhanced cursor validation, which let cursor can be fetched after transaction committed if it has not read uncommitted data. | False |
| _ob_query_rate_limit | TENANT | the maximun throughput allowed for a tenant per observer instance | -1 |
| _max_schema_slot_num | TENANT | the max schema slot number for multi-version schema memory management, Range: [2, 256] in integer | 128 |
| _audit_mode | TENANT | specifies audit mode,NONE: close audit,MYSQL: use mysql auditORACLE: use oracle audit | NONE |
| _bloom_filter_enabled | TENANT | enable join bloom filter | True |
| _backup_task_keep_alive_timeout | TENANT | control backup task keep alive timeoutRange: [1s, +∞) | 10m |
| _backup_task_keep_alive_interval | TENANT | control backup task keep alive intervalRange: [1s, +∞) | 10s |
| _ob_ddl_temp_file_compress_func | TENANT | specific compression in ObTempBlockStore.AUTO: use compression algorithm from table schema;ZSTD: use ZSTD compression algorithm;LZ4: use LZ4 compression algorithm;NONE: do not use compression. | AUTO |
| _max_ls_cnt_per_server | TENANT | specify max ls count of one tenant on one observer.WARNING: Modifying this can potentially destabilize the cluster. It is strongly advised to avoid making such changes as they are unlikely to be necessary.0: the cluster will adapt the max ls number according to the memory size of tenant itself | 0 |
| _ob_enable_fast_freeze | TENANT | specifies whether the tenant's fast freeze is enabledValue: True:turned on; False: turned off | True |
| _minor_compaction_amplification_factor | TENANT | thre L1 compaction write amplification factor, 0 means default 25, Range: [0,100] in integer | 0 |
| _trx_max_log_cb_limit | TENANT | Control the upper limit of TxLogCbs involved in the participant to manage the maximum concurrency of submiting logs in a transaction | 16 |
| _enable_adaptive_compaction | TENANT | specifies whether allow adaptive compaction schedule and information collection | True |
| _enable_parallel_minor_merge | TENANT | specifies whether enable parallel minor merge. Value: True:turned on; False: turned off | True |
| _io_callback_thread_count | TENANT | The number of io callback threads. The default value is 8. Range: [1,64] in integer | 8 |
| _stream_rpc_max_wait_timeout | TENANT | the maximum timeout for a tenant worker thread to wait for the next request while processing streaming RPC | 30s |
| _keepalive_interval | TENANT | This configuration item is used to control the write interval of the keepalive log, ensuring that the readable point of the standby database can be advanced in a timely manner. | 100ms |
| _tx_result_retention | TENANT | The tx data can be recycled after at least _tx_result_retention seconds. Range: [0, 36000] | 300 |
| _ob_enable_standby_db_parallel_log_transport | TENANT | Specifies whether the parallel log transport protocol is enabled on the standby database. The parallel log transport protocol is enabled only if this parameter is true and the primary database is compatible with the parallel log transport protocol. | True |
| _ls_gc_wait_readonly_tx_time | TENANT | The maximum waiting time for residual read-only transaction before executing log stream garbage collecting。The default value is 24h. Range: [0s, +∞).Log stream garbage collecting will no longer wait for readonly transaction when the tenant is dropped. | 24h |
| _log_writer_parallelism | TENANT | the number of parallel log writer threads that can be used to write redo log entries to disk. | 3 |
| _parallel_ddl_control | TENANT | switch for parallel capability of parallel DDL | |
| _enable_parallel_table_creation | TENANT | specifies whether create table parallelly. Value: True: create table parallelly; False: create table serially | True |
| _max_tablet_cnt_per_gb | TENANT | The maximum number of tablets supported per 1GB of memory by tenant unit. Range: [1000, 50000) | 20000 |
| _storage_meta_memory_limit_percentage | TENANT | maximum memory for storage meta, as a percentage of total tenant memory. Range: [0, 50), percentage, 0 means no limit to storage meta memory | 20 |
| _temporary_file_meta_memory_limit_percentage | TENANT | The memory limit of temporary file meta, and the value is a percentage of the tenant's memory. The default value is 70. For compatibility, 0 is 70% of tenant memory.Range: [0, 100], percentage | 0 |
| _temporary_file_io_area_size | TENANT | memory buffer size of temporary file, as a percentage of total tenant memory. Range: [0, 50), percentage | 1 |
| _pushdown_storage_level | TENANT | the level of storage pushdown. Range: [0, 3] 0: disabled, 1:blockscan, 2: blockscan & filter, 3: blockscan & filter & aggregate | 3 |
| _enable_hash_join_processor | TENANT | which path to process for hash join, default 7 to auto choose 1: nest loop, 2: recursive, 4: in-memory | 7 |
| _enable_hash_join_hasher | TENANT | which hash function to choose for hash join 1: murmurhash, 2: crc, 4: xxhash | 1 |
| _force_hash_join_spill | TENANT | force hash join to dump after get all build hash table Value: True:turned on False: turned off | False |
| _force_hash_groupby_dump | TENANT | force hash groupby to dumpValue: True:turned on False: turned off | False |
| _sqlexec_disable_hash_based_distagg_tiv | TENANT | disable hash based distinct aggregation in the second stage of three stage aggregation for gby queriesValue: True:turned on False: turned off | False |
| _px_message_compression | TENANT | Enable DTL send message with compressionValue: True: enable compression False: disable compression | True |
| _px_max_message_pool_pct | TENANT | The maxinum percent of tenant memory that DTL message buffer pool can alloc memory. Range: [0,90] | 40 |
| _parallel_server_sleep_time | TENANT | sleep time between get channel data in millisecond. Range: [0, 2000] | 1 |
| _parallel_min_message_pool | TENANT | DTL message buffer pool reserve the mininum size after extend the size. Range: [16M,8G] | 16M |
| _mds_memory_limit_percentage | TENANT | Used to control the upper limit percentage of memory resources that the Mds module can use. Range:(0, 100) | 10 |
| _tx_data_memory_limit_percentage | TENANT | used to control the upper limit percentage of memory resources that the TxData module can use. Range:(0, 100) | 20 |
| _tx_share_memory_limit_percentage | TENANT | Used to control the percentage of tenant memory limit that multiple modules in the transaction layer can collectively use. This primarily includes user data (MemTable), transaction data (TxData), etc. When it is set to the default value of 0, it represents dynamic adaptive behavior, which will be adjusted dynamically based on memstore_limit_percentage. The adjustment rule is: _tx_share_memory_limit_percentage = memstore_limit_percentage + 10. Range: [0, 100) | 0 |
| _memstore_limit_percentage | TENANT | used in calculating the value of MEMSTORE_LIMIT parameter: _memstore_limit_percentage = memstore_limit / memory_size, where MEMORY_SIZE is determined when the tenant is created. Range: [0, 100). 1. the system will use memstore_limit_percentage if only memstore_limit_percentage is set.2. the system will use _memstore_limit_percentage if both memstore_limit_percentage and _memstore_limit_percentage is set.3. the system will adjust automatically if both memstore_limit_percentage and _memstore_limit_percentage set to 0(by default). | 0 |
| _enable_enum_set_subschema | TENANT | Specifies whether to enable the enum/set extended type info is stored as subschema and to activate the related new type cast logic behavior. | True |
| _enable_mysql_compatible_dates | TENANT | Specifies whether to use MySQL-compatible date format that allows for invalid dates. | True |
| _adaptive_join_enabled | TENANT | enable/disable adaptive join | False |
| _nested_loop_join_enabled | TENANT | enable/disable nested loop join | True |
| _optimizer_sortmerge_join_enabled | TENANT | enable/disable merge join | True |
| _hash_join_enabled | TENANT | enable/disable hash join | True |
| _optimizer_ads_time_limit | TENANT | the maximum optimizer dynamic sampling time limit. Range: [0, 300] | 10 |
| _ob_enable_dynamic_worker | TENANT | specifies whether worker count increases when all workers were in blocking. | True |
| _enable_convert_real_to_decimal | TENANT | specifies whether convert column type float(M,D), double(M,D) to decimal(M,D) in DDL | False |
| _ctx_memory_limit | TENANT | specifies tenant ctx memory limit. | |
| _rowsets_max_rows | TENANT | the row number processed by vectorized sql engine within one batch. Range: [0, 65535] | 256 |
| _rowsets_target_maxsize | TENANT | the size of the memory reserved for vectorized sql engine. Range: [262144, 8388608] | 524288 |
| _rowsets_enabled | TENANT | specifies whether vectorized sql execution engine is activated | True |
| _mvcc_gc_using_min_txn_snapshot | TENANT | specifies enable mvcc gc using active txn snapshot | True |
| _hash_area_size | TENANT | size of maximum memory that could be used by HASH JOIN. Range: [4M,+∞) | 32M |
| _sort_area_size | TENANT | size of maximum memory that could be used by SORT. Range: [2M,+∞) | 32M |
| _publish_schema_mode | TENANT | specify the inspection of schema synchronous status after ddl transaction commitsvalues: BEST_EFFORT, ASYNC | BEST_EFFORT |
+-------------------------------------------------+---------+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+------------------------------------------------------------------------------------+
226 rows in set (0.117 sec)