| Category | +Related documents | +
|---|---|
| TiDB architecture and principles | +TiDB Architecture FAQs | +
| Deployment | ++ |
| Data migration | +
+
|
+
| Data backup and restore | +Backup & Restore FAQs | +
| SQL operations | +SQL FAQs | +
| Cluster upgrade | +TiDB Upgrade FAQs | +
| Cluster management | +Cluster Management FAQs | +
| Monitor and alert | ++ |
| High availability and high reliability | ++ |
| Common error codes | +Error Codes and Troubleshooting | +
| Operating systems | +Supported CPU architectures | +
|---|---|
| Red Hat Enterprise Linux 8.4 or a later 8.x version | +
|
+
|
+
|
+
| Amazon Linux 2 | +
|
+
| Kylin Euler V10 SP1/SP2 | +
|
+
| UOS V20 | +
|
+
| openEuler 22.03 LTS SP1 | +x86_64 | +
| macOS 12 (Monterey) or later | +
|
+
| Oracle Enterprise Linux 7.3 or a later 7.x version | +x86_64 | +
| Ubuntu LTS 18.04 or later | +x86_64 | +
| CentOS 8 Stream | +
|
+
| Debian 9 (Stretch) or later | +x86_64 | +
| Fedora 35 or later | +x86_64 | +
| openSUSE Leap later than v15.3 (not including Tumbleweed) | +x86_64 | +
| SUSE Linux Enterprise Server 15 | +x86_64 | +
| Component | +Disk space requirement | +Healthy disk usage | +
|---|---|---|
| TiDB | +
|
+ Lower than 90% | +
| PD | +At least 20 GB for the data disk and for the log disk, respectively | +Lower than 90% | +
| TiKV | +At least 100 GB for the data disk and for the log disk, respectively | +Lower than 80% | +
| TiFlash | +At least 100 GB for the data disk and at least 30 GB for the log disk, respectively | +Lower than 80% | +
| TiUP | +
|
+ N/A | +
| Ngmonitoring | +
|
+ N/A | +
| Variable name | +Change type | +Description | +
|---|---|---|
tidb_enable_column_tracking |
+ Newly added | +Controls whether to allow TiDB to collect PREDICATE COLUMNS. The default value is OFF. |
+
tidb_enable_paging |
+ Newly added | +Controls whether to use the method of paging to send coprocessor requests in IndexLookUp operator. The default value is OFF.For read queries that use IndexLookup and Limit and that Limit cannot be pushed down to IndexScan, there might be high latency for the read queries and high CPU usage for TiKV's unified read pool. In such cases, because the Limit operator only requires a small set of data, if you set tidb_enable_paging to ON, TiDB processes less data, which reduces query latency and resource consumption. |
+
tidb_enable_top_sql |
+ Newly added | +Controls whether to enable the Top SQL feature. The default value is OFF. |
+
tidb_persist_analyze_options |
+ Newly added | +Controls whether to enable the ANALYZE configuration persistence feature. The default value is ON. |
+
tidb_read_staleness |
+ Newly added | +Controls the range of historical data that can be read in the current session. The default value is 0. |
+
tidb_regard_null_as_point |
+ Newly added | +Controls whether the optimizer can use a query condition including null equivalence as a prefix condition for index access. | +
tidb_stats_load_sync_wait |
+ Newly added | +Controls whether to enable the synchronously loading statistics feature. The default value 0 means that the feature is disabled and that the statistics is asynchronously loaded. When the feature is enabled, this variable controls the maximum time that SQL optimization can wait for synchronously loading statistics before timeout. |
+
tidb_stats_load_pseudo_timeout |
+ Newly added | +Controls when synchronously loading statistics reaches timeout, whether SQL fails (OFF) or falls back to using pseudo statistics (ON). The default value is OFF. |
+
tidb_backoff_lock_fast |
+ Modified | +The default value is changed from 100 to 10. |
+
tidb_enable_index_merge |
+ Modified | +The default value is changed from OFF to ON.
|
+
tidb_store_limit |
+ Modified | +Before v5.4.0, this variable can be configured at instance level and globally. Starting from v5.4.0, this variable only supports global configuration. | +
| Variable name | +Change type | +Description | +
|---|---|---|
placement_checks |
+ Deleted | +Controls whether the DDL statement validates the placement rules specified by Placement Rules in SQL. Replaced by tidb_placement_mode. |
+
tidb_enable_alter_placement |
+ Deleted | +Controls whether to enable placement rules in SQL. | +
+ tidb_mem_quota_hashjoin+ tidb_mem_quota_indexlookupjoin+ tidb_mem_quota_indexlookupreader+ tidb_mem_quota_mergejoin+ tidb_mem_quota_sort+ tidb_mem_quota_topn
+ |
+ Deleted | +Since v5.0, these variables have been replaced by tidb_mem_quota_query and removed from the system variables document. To ensure compatibility, these variables were kept in source code. Since TiDB 6.0.0, these variables are removed from the code, too. |
+
tidb_enable_mutation_checker |
+ Newly added | +Controls whether to enable the mutation checker. The default value is ON. For existing clusters that upgrade from versions earlier than v6.0.0, the mutation checker is disabled by default. |
+
tidb_ignore_prepared_cache_close_stmt |
+ Newly added | +Controls whether to ignore the command that closes Prepared Statement. The default value is OFF. |
+
tidb_mem_quota_binding_cache |
+ Newly added | +Sets the memory usage threshold for the cache holding binding. The default value is 67108864 (64 MiB). |
+
tidb_placement_mode |
+ Newly added | +Controls whether DDL statements ignore the placement rules specified by Placement Rules in SQL. The default value is strict, which means that DDL statements do not ignore placement rules. |
+
tidb_rc_read_check_ts |
+ Newly added | +
+
|
+
tidb_sysdate_is_now |
+ Newly added | +Controls whether the SYSDATE function can be replaced by the NOW function. This configuration item has the same effect as the MySQL option sysdate-is-now. The default value is OFF. |
+
tidb_table_cache_lease |
+ Newly added | +Controls the lease time of table cache, in seconds. The default value is 3. |
+
tidb_top_sql_max_meta_count |
+ Newly added | +Controls the maximum number of SQL statement types collected by Top SQL per minute. The default value is 5000. |
+
tidb_top_sql_max_time_series_count |
+ Newly added | +Controls how many SQL statements that contribute the most to the load (that is, top N) can be recorded by Top SQL per minute. The default value is 100. |
+
tidb_txn_assertion_level |
+ Newly added | +Controls the assertion level. The assertion is a consistency check between data and indexes, which checks whether a key being written exists in the transaction commit process. By default, the check enables most of the check items, with almost no impact on performance. For existing clusters that upgrade from versions earlier than v6.0.0, the check is disabled by default. | +
KB\|MB\|GB.| Configuration file | +Configuration | +Change type | +Description | +
|---|---|---|---|
| TiDB | +
+ stmt-summary.enable+ stmt-summary.enable-internal-query+ stmt-summary.history-size+ stmt-summary.max-sql-length+ stmt-summary.max-stmt-count+ stmt-summary.refresh-interval
+ |
+ Deleted | +Configuration related to the statement summary tables. All these configuration items are removed. You need to use SQL variables to control the statement summary tables. | +
| TiDB | +new_collations_enabled_on_first_bootstrap |
+ Modified | +Controls whether to enable support for the new collation. Since v6.0, the default value is changed from false to true. This configuration item only takes effect when the cluster is initialized for the first time. After the first bootstrap, you cannot enable or disable the new collation framework using this configuration item. |
+
| TiKV | +backup.num-threads |
+ Modified | +The value range is modified to [1, CPU]. |
+
| TiKV | +raftstore.apply-max-batch-size |
+ Modified | +The maximum value is changed to 10240. |
+
| TiKV | +raftstore.raft-max-size-per-msg |
+ Modified | +The minimum value is changed from 0 to larger than 0.The maximum value is set to 3GB.The unit is changed from MB to KB|MB|GB. |
+
| TiKV | +raftstore.store-max-batch-size |
+ Modified | +The maximum value is set to 10240. |
+
| TiKV | +readpool.unified.max-thread-count |
+ Modified | +The adjustable range is changed to [min-thread-count, MAX(4, CPU)]. |
+
| TiKV | +rocksdb.enable-pipelined-write |
+ Modified | +The default value is changed from true to false. When this configuration is enabled, the previous Pipelined Write is used. When this configuration is disabled, the new Pipelined Commit mechanism is used. |
+
| TiKV | +rocksdb.max-background-flushes |
+ Modified | +When the number of CPU cores is 10, the default value is 3.When the number of CPU cores is 8, the default value is 2. |
+
| TiKV | +rocksdb.max-background-jobs |
+ Modified | +When the number of CPU cores is 10, the default value is 9.When the number of CPU cores is 8, the default value is 7. |
+
| TiFlash | +profiles.default.dt_enable_logical_split |
+ Modified | +Determines whether the segment of DeltaTree Storage Engine uses logical split. The default value is changed from true to false. |
+
| TiFlash | +profiles.default.enable_elastic_threadpool |
+ Modified | +Controls whether to enable the elastic thread pool. The default value is changed from false to true. |
+
| TiFlash | +storage.format_version |
+ Modified | +Controls the data validation feature of TiFlash. The default value is changed from 2 to 3.When format_version is set to 3, consistency check is performed on the read operations for all TiFlash data to avoid incorrect read due to hardware failure.Note that the new format version cannot be downgraded in place to versions earlier than v5.4. |
+
| TiDB | +pessimistic-txn.pessimistic-auto-commit |
+ Newly added | +Determines the transaction mode that the auto-commit transaction uses when the pessimistic transaction mode is globally enabled (tidb_txn_mode='pessimistic'). |
+
| TiKV | +pessimistic-txn.in-memory |
+ Newly added | +Controls whether to enable the in-memory pessimistic lock. With this feature enabled, pessimistic transactions store pessimistic locks in TiKV memory as much as possible, instead of writing pessimistic locks to disks or replicating to other replicas. This improves the performance of pessimistic transactions; however, there is a low probability that a pessimistic lock will be lost, which might cause the pessimistic transaction to fail to commit. The default value is true. |
+
| TiKV | +quota |
+ Newly added | +Add configuration items related to Quota Limiter, which limit the resources occupied by frontend requests. Quota Limiter is an experimental feature and is disabled by default. New quota-related configuration items are foreground-cpu-time, foreground-write-bandwidth, foreground-read-bandwidth, and max-delay-duration. |
+
| TiFlash | +profiles.default.dt_compression_method |
+ Newly added | +Specifies the compression algorithm for TiFlash. The optional values are LZ4, zstd and LZ4HC, all case insensitive. The default value is LZ4. |
+
| TiFlash | +profiles.default.dt_compression_level |
+ Newly added | +Specifies the compression level of TiFlash. The default value is 1. |
+
| DM | +loaders.<name>.import-mode |
+ Newly added | +The import mode during the full import phase. Since v6.0, DM uses TiDB Lightning's TiDB-backend mode to import data during the full import phase; the previous Loader component is no longer used. This is an internal replacement and has no obvious impact on daily operations. The default value is set to sql, which means using tidb-backend mode. In some rare cases, tidb-backend might not be fully compatible. You can fall back to Loader mode by configuring this parameter to loader. |
+
| DM | +loaders.<name>.on-duplicate |
+ Newly added | +Specifies the methods to resolve conflicts during the full import phase. The default value is replace, which means using the new data to replace the existing data. |
+
| TiCDC | +dial-timeout |
+ Newly added | +The timeout in establishing a connection with the downstream Kafka. The default value is 10s. |
+
| TiCDC | +read-timeout |
+ Newly added | +The timeout in getting a response returned by the downstream Kafka. The default value is 10s. |
+
| TiCDC | +write-timeout |
+ Newly added | +The timeout in sending a request to the downstream Kafka. The default value is 10s. |
+