TiDB Configuration File
The TiDB configuration file supports more options than command-line parameters. You can download the default configuration file config.toml.example
and rename it to config.toml
. This document describes only the options that are not involved in command line options.
split-table
- Determines whether to create a separate Region for each table.
- Default value:
true
- It is recommended to set it to
false
if you need to create a large number of tables (for example, more than 100 thousand tables).
tidb-max-reuse-chunk
New in v6.4.0
- Controls the maximum cached chunk objects of chunk allocation. Setting this configuration item to too large a value might increase the risk of OOM.
- Default value:
64
- Minimum value:
0
- Maximum value:
2147483647
tidb-max-reuse-column
New in v6.4.0
- Controls the maximum cached column objects of chunk allocation. Setting this configuration item to too large a value might increase the risk of OOM.
- Default value:
256
- Minimum value:
0
- Maximum value:
2147483647
token-limit
- The number of sessions that can execute requests concurrently.
- Type: Integer
- Default value:
1000
- Minimum value:
1
- Maximum value:
1048576
temp-dir
New in v6.3.0
- File system location used by TiDB to store temporary data. If a feature requires local storage in TiDB nodes, TiDB stores the corresponding temporary data in this location.
- When creating an index, if
tidb_ddl_enable_fast_reorg
is enabled, data that needs to be backfilled for a newly created index will be at first stored in the TiDB local temporary directory, and then imported into TiKV in batches, thus accelerating the index creation. - When
IMPORT INTO
is used to import data, the sorted data is first stored in the TiDB local temporary directory, and then imported into TiKV in batches. - Default value:
"/tmp/tidb"
oom-use-tmp-storage
- Controls whether to enable the temporary storage for some operators when a single SQL statement exceeds the memory quota specified by the system variable
tidb_mem_quota_query
. - Default value:
true
tmp-storage-path
- Specifies the temporary storage path for some operators when a single SQL statement exceeds the memory quota specified by the system variable
tidb_mem_quota_query
. - Default value:
<temporary directory of OS>/<OS user ID>_tidb/MC4wLjAuMDo0MDAwLzAuMC4wLjA6MTAwODA=/tmp-storage
.MC4wLjAuMDo0MDAwLzAuMC4wLjA6MTAwODA=
is theBase64
encoding result of<host>:<port>/<statusHost>:<statusPort>
. - This configuration takes effect only when the system variable
tidb_enable_tmp_storage_on_oom
isON
.
tmp-storage-quota
- Specifies the quota for the storage in
tmp-storage-path
. The unit is byte. - When a single SQL statement uses a temporary disk and the total volume of the temporary disk of the TiDB server exceeds this configuration value, the current SQL operation is cancelled and the
Out of Global Storage Quota!
error is returned. - When the value of this configuration is smaller than
0
, the above check and limit do not apply. - Default value:
-1
- When the remaining available storage in
tmp-storage-path
is lower than the value defined bytmp-storage-quota
, the TiDB server reports an error when it is started, and exits.
lease
- The timeout of the DDL lease.
- Default value:
45s
- Unit: second
compatible-kill-query
- Determines whether to set the
KILL
statement to be MySQL compatible. - Default value:
false
compatible-kill-query
takes effect only whenenable-global-kill
is set tofalse
.- When
enable-global-kill
isfalse
,compatible-kill-query
controls whether you need to append theTIDB
keyword when killing a query.- When
compatible-kill-query
isfalse
, the behavior ofKILL xxx
in TiDB is different from that in MySQL. To kill a query in TiDB, you need to append theTIDB
keyword, such asKILL TIDB xxx
. - When
compatible-kill-query
istrue
, to kill a query in TiDB, there is no need to append theTIDB
keyword. It is STRONGLY NOT RECOMMENDED to setcompatible-kill-query
totrue
in your configuration file UNLESS you are certain that clients will be always connected to the same TiDB instance. This is because pressing Control+C in the default MySQL client opens a new connection in whichKILL
is executed. If there is a proxy between the client and the TiDB cluster, the new connection might be routed to a different TiDB instance, which possibly kills a different session by mistake.
- When
- When
enable-global-kill
istrue
,KILL xxx
andKILL TIDB xxx
have the same effect. - For more information about the
KILL
statement, see KILL [TIDB].
check-mb4-value-in-utf8
- Determines whether to enable the
utf8mb4
character check. When this feature is enabled, if the character set isutf8
and themb4
characters are inserted inutf8
, an error is returned. - Default value:
false
- Since v6.1.0, whether to enable the
utf8mb4
character check is determined by the TiDB configuration iteminstance.tidb_check_mb4_value_in_utf8
or the system variabletidb_check_mb4_value_in_utf8
.check-mb4-value-in-utf8
still takes effect. But if bothcheck-mb4-value-in-utf8
andinstance.tidb_check_mb4_value_in_utf8
are set, the latter takes effect.
treat-old-version-utf8-as-utf8mb4
- Determines whether to treat the
utf8
character set in old tables asutf8mb4
. - Default value:
true
alter-primary-key
(Deprecated)
- Determines whether to add or remove the primary key constraint to or from a column.
- Default value:
false
- With this default setting, adding or removing the primary key constraint is not supported. You can enable this feature by setting
alter-primary-key
totrue
. However, if a table already exists before the switch is on, and the data type of its primary key column is an integer, dropping the primary key from the column is not possible even if you set this configuration item totrue
.
server-version
- Modifies the version string returned by TiDB in the following situations:
- When the built-in
VERSION()
function is used. - When TiDB establishes the initial connection to the client and returns the initial handshake packet with version string of the server. For details, see MySQL Initial Handshake Packet.
- When the built-in
- Default value: ""
- By default, the format of the TiDB version string is
8.0.11-TiDB-${tidb_version}
.
repair-mode
- Determines whether to enable the untrusted repair mode. When the
repair-mode
is set totrue
, bad tables in therepair-table-list
cannot be loaded. - Default value:
false
- The
repair
syntax is not supported by default. This means that all tables are loaded when TiDB is started.
repair-table-list
repair-table-list
is only valid whenrepair-mode
is set totrue
.repair-table-list
is a list of bad tables that need to be repaired in an instance. An example of the list is: ["db.table1","db.table2"...].- Default value: []
- The list is empty by default. This means that there are no bad tables that need to be repaired.
new_collations_enabled_on_first_bootstrap
- Enables or disables the new collation support.
- Default value:
true
- Note: This configuration takes effect only for the TiDB cluster that is first initialized. After the initialization, you cannot use this configuration item to enable or disable the new collation support.
max-server-connections
- The maximum number of concurrent client connections allowed in TiDB. It is used to control resources.
- Default value:
0
- By default, TiDB does not set limit on the number of concurrent client connections. When the value of this configuration item is greater than
0
and the number of actual client connections reaches this value, the TiDB server rejects new client connections. - Since v6.2.0, the TiDB configuration item
instance.max_connections
or the system variablemax_connections
is used to set the maximum number of concurrent client connections allowed in TiDB.max-server-connections
still takes effect. But ifmax-server-connections
andinstance.max_connections
are set at the same time, the latter takes effect.
max-index-length
- Sets the maximum allowable length of the newly created index.
- Default value:
3072
- Unit: byte
- Currently, the valid value range is
[3072, 3072*4]
. MySQL and TiDB (version < v3.0.11) do not have this configuration item, but both limit the length of the newly created index. This limit in MySQL is3072
. In TiDB (version =< 3.0.7), this limit is3072*4
. In TiDB (3.0.7 < version < 3.0.11), this limit is3072
. This configuration is added to be compatible with MySQL and earlier versions of TiDB.
table-column-count-limit
New in v5.0
- Sets the limit on the number of columns in a single table.
- Default value:
1017
- Currently, the valid value range is
[1017, 4096]
.
index-limit
New in v5.0
- Sets the limit on the number of indexes in a single table.
- Default value:
64
- Currently, the valid value range is
[64, 512]
.
enable-telemetry
New in v4.0.2 and deprecated in v8.1.0
- Before v8.1.0, this configuration item controls whether to enable telemetry collection in a TiDB instance.
- Default value:
false
deprecate-integer-display-length
- Deprecates the display width for integer types when this configuration item is set to
true
. - Default value:
false
enable-tcp4-only
New in v5.0
- Enables or disables listening on TCP4 only.
- Default value:
false
- Enabling this option is useful when TiDB is used with LVS for load balancing because the real client IP from the TCP header can be correctly parsed by the "tcp4" protocol.
enable-enum-length-limit
New in v5.0
- Determines whether to limit the maximum length of a single
ENUM
element and a singleSET
element. - Default value:
true
- When this configuration value is
true
, the maximum length of a singleENUM
element and a singleSET
element is 255 characters, which is compatible with MySQL 8.0. When this configuration value isfalse
, there is no limit on the length of a single element, which is compatible with TiDB (earlier than v5.0).
graceful-wait-before-shutdown
New in v5.0
- Specifies the number of seconds that TiDB waits when you shut down the server, which allows the clients to disconnect.
- Default value:
0
- When TiDB is waiting for shutdown (in the grace period), the HTTP status will indicate a failure, which allows the load balancers to reroute traffic.
enable-global-kill
New in v6.1.0
- Controls whether to enable the Global Kill (terminating queries or connections across instances) feature.
- Default value:
true
- When the value is
true
, bothKILL
andKILL TIDB
statements can terminate queries or connections across instances so you do not need to worry about erroneously terminating queries or connections. When you use a client to connect to any TiDB instance and execute theKILL
orKILL TIDB
statement, the statement will be forwarded to the target TiDB instance. If there is a proxy between the client and the TiDB cluster, theKILL
andKILL TIDB
statements will also be forwarded to the target TiDB instance for execution. - Starting from v7.3.0, you can terminate a query or connection using the MySQL command line Control+C when both
enable-global-kill
andenable-32bits-connection-id
are set totrue
. For more information, seeKILL
.
enable-32bits-connection-id
New in v7.3.0
- Controls whether to enable the 32-bit connection ID feature.
- Default value:
true
- When both this configuration item and
enable-global-kill
are set totrue
, TiDB generates 32-bit connection IDs. This enables you to terminate queries or connections by the MySQL command-line Control+C.
initialize-sql-file
New in v6.6.0
- Specifies the SQL script to be executed when the TiDB cluster is started for the first time.
- Default value:
""
- All SQL statements in this script are executed with the highest privilege without any privilege check. If the specified SQL script fails to execute, the TiDB cluster might fail to start.
- This configuration item is used to perform such operations as modifying the value of a system variable, creating a user, or granting privileges.
enable-forwarding
New in v5.0.0
- Controls whether the PD client and TiKV client in TiDB forward requests to the leader via the followers in the case of possible network isolation.
- Default value:
false
- If the environment might have isolated network, enabling this parameter can reduce the window of service unavailability.
- If you cannot accurately determine whether isolation, network interruption, or downtime has occurred, using this mechanism has the risk of misjudgment and causes reduced availability and performance. If network failure has never occurred, it is not recommended to enable this parameter.
enable-table-lock
New in v4.0.0
- Controls whether to enable the table lock feature.
- Default value:
false
- The table lock is used to coordinate concurrent access to the same table among multiple sessions. Currently, the
READ
,WRITE
, andWRITE LOCAL
lock types are supported. When the configuration item is set tofalse
, executing theLOCK TABLES
orUNLOCK TABLES
statement does not take effect and returns the "LOCK/UNLOCK TABLES is not supported" warning. For more information, seeLOCK TABLES
andUNLOCK TABLES
.
labels
- Specify server labels. For example,
{ zone = "us-west-1", dc = "dc1", rack = "rack1", host = "tidb1" }
. - Default value:
{}
Log
Configuration items related to log.
level
- Specifies the log output level.
- Value options:
debug
,info
,warn
,error
, andfatal
. - Default value:
info
format
- Specifies the log output format.
- Value options:
json
andtext
. - Default value:
text
enable-timestamp
- Determines whether to enable timestamp output in the log.
- Default value:
null
- If you set the value to
false
, the log does not output timestamp.
enable-slow-log
- Determines whether to enable the slow query log.
- Default value:
true
- To enable the slow query log, set
enable-slow-log
totrue
. Otherwise, set it tofalse
. - Since v6.1.0, whether to enable slow query log is determined by the TiDB configuration item
instance.tidb_enable_slow_log
or the system variabletidb_enable_slow_log
.enable-slow-log
still takes effect. But ifenable-slow-log
andinstance.tidb_enable_slow_log
are set at the same time, the latter takes effect.
slow-query-file
- The file name of the slow query log.
- Default value:
tidb-slow.log
- The format of the slow log is updated in TiDB v2.1.8, so the slow log is output to the slow log file separately. In versions before v2.1.8, this variable is set to "" by default.
- After you set it, the slow query log is output to this file separately.
slow-threshold
- Outputs the threshold value of consumed time in the slow log.
- Default value:
300
- Unit: Milliseconds
- When the time consumed by a query is larger than this value, this query is considered as a slow query and its log is output to the slow query log. Note that when the output level of
log.level
is"debug"
, all queries are recorded in the slow query log, regardless of the setting of this parameter. - Since v6.1.0, the threshold value of consumed time in the slow log is specified by the TiDB configuration item
instance.tidb_slow_log_threshold
or the system variabletidb_slow_log_threshold
.slow-threshold
still takes effect. But ifslow-threshold
andinstance.tidb_slow_log_threshold
are set at the same time, the latter takes effect.
record-plan-in-slow-log
- Determines whether to record execution plans in the slow log.
- Default value:
1
- Since v6.1.0, whether to record execution plans in the slow log is determined by the TiDB configuration item
instance.tidb_record_plan_in_slow_log
or the system variabletidb_record_plan_in_slow_log
.record-plan-in-slow-log
still takes effect. But ifrecord-plan-in-slow-log
andinstance.tidb_record_plan_in_slow_log
are set at the same time, the latter takes effect.
expensive-threshold
- Outputs the threshold value of the number of rows for the
expensive
operation. - Default value:
10000
- When the number of query rows (including the intermediate results based on statistics) is larger than this value, it is an
expensive
operation and outputs log with the[EXPENSIVE_QUERY]
prefix.
general-log-file
New in v8.0.0
- The filename of the general log.
- Default value:
""
- If you specify a filename, the general log is written to this specified file. If the value is blank, the general log is written to the server log of the TiDB instance. You can specify the name of the server log using
filename
.
timeout
New in v7.1.0
- Sets the timeout for log-writing operations in TiDB. In case of a disk failure that prevents logs from being written, this configuration item can trigger the TiDB process to panic instead of hang.
- Default value:
0
, indicating no timeout is set. - Unit: second
- In some user scenarios, TiDB logs might be stored on hot-pluggable or network-attached disks, which might become permanently unavailable. In these cases, TiDB cannot recover automatically from such disaster and the log-writing operations will be permanently blocked. Although the TiDB process might seem to be running, it does not respond to any requests. This configuration item is designed to handle such situations.
log.file
Configuration items related to log files.
filename
- The file name of the general log file.
- Default value: ""
- If you set it, the log is output to this file.
max-size
- The size limit of the log file.
- Default value: 300
- Unit: MB
- The maximum value is 4096.
max-days
- The maximum number of days that the log is retained.
- Default value:
0
- The log is retained by default. If you set the value, the expired log is cleaned up after
max-days
.
max-backups
- The maximum number of retained logs.
- Default value:
0
- All the log files are retained by default. If you set it to
7
, seven log files are retained at maximum.
compression
New in v8.0.0
- The compression method for the log.
- Default value:
""
- Value options:
""
,"gzip"
- The default value is
""
, which means no compression. To enable the gzip compression, set this value to"gzip"
. After compression is enabled, all log files are affected, such asslow-query-file
andgeneral-log-file
.
Security
Configuration items related to security.
enable-sem
- Enables the Security Enhanced Mode (SEM).
- Default value:
false
- The status of SEM is available via the system variable
tidb_enable_enhanced_security
.
ssl-ca
- The file path of the trusted CA certificate in the PEM format.
- Default value: ""
- If you set this option and
--ssl-cert
,--ssl-key
at the same time, TiDB authenticates the client certificate based on the list of trusted CAs specified by this option when the client presents the certificate. If the authentication fails, the connection is terminated. - If you set this option but the client does not present the certificate, the secure connection continues without client certificate authentication.
ssl-cert
- The file path of the SSL certificate in the PEM format.
- Default value: ""
- If you set this option and
--ssl-key
at the same time, TiDB allows (but not forces) the client to securely connect to TiDB using TLS. - If the specified certificate or private key is invalid, TiDB starts as usual but cannot receive secure connection.
ssl-key
- The file path of the SSL certificate key in the PEM format, that is, the private key of the certificate specified by
--ssl-cert
. - Default value: ""
- Currently, TiDB does not support loading the private keys protected by passwords.
cluster-ssl-ca
- The CA root certificate used to connect TiKV or PD with TLS.
- Default value: ""
cluster-ssl-cert
- The path of the SSL certificate file used to connect TiKV or PD with TLS.
- Default value: ""
cluster-ssl-key
- The path of the SSL private key file used to connect TiKV or PD with TLS.
- Default value: ""
spilled-file-encryption-method
- Determines the encryption method used for saving the spilled files to disk.
- Default value:
"plaintext"
, which disables encryption. - Optional values:
"plaintext"
and"aes128-ctr"
auto-tls
- Determines whether to automatically generate the TLS certificates on startup.
- Default value:
false
tls-version
- Set the minimum TLS version for MySQL Protocol connections.
- Default value: "", which allows TLSv1.2 or later versions. Before TiDB v7.6.0, the default value allows TLSv1.1 or later versions.
- Optional values:
"TLSv1.2"
and"TLSv1.3"
. Before TiDB v8.0.0,"TLSv1.0"
and"TLSv1.1"
are also allowed.
auth-token-jwks
New in v6.4.0
- Set the local file path of the JSON Web Key Sets (JWKS) for the
tidb_auth_token
authentication method. - Default value:
""
auth-token-refresh-interval
New in v6.4.0
- Set the JWKS refresh interval for the
tidb_auth_token
authentication method. - Default value:
1h
disconnect-on-expired-password
New in v6.5.0
- Determines whether TiDB disconnects the client connection when the password is expired.
- Default value:
true
- Optional values:
true
,false
- If you set it to
true
, the client connection is disconnected when the password is expired. If you set it tofalse
, the client connection is restricted to the "sandbox mode" and the user can only execute the password reset operation.
session-token-signing-cert
New in v6.4.0
- The certificate file path, which is used by TiProxy for session migration.
- Default value: ""
- Empty value will cause TiProxy session migration to fail. To enable session migration, all TiDB nodes must set this to the same certificate and key. This means that you should store the same certificate and key on every TiDB node.
session-token-signing-key
New in v6.4.0
- The key file path used by TiProxy for session migration.
- Default value: ""
- Refer to the descriptions of
session-token-signing-cert
.
Performance
Configuration items related to performance.
max-procs
- The number of CPUs used by TiDB.
- Default value:
0
- The default
0
indicates using all the CPUs on the machine. You can also set it to n, and then TiDB uses n CPUs.
server-memory-quota
New in v4.0.9
- The memory usage limit of tidb-server instances.
- Default value:
0
(in bytes), which means no memory limit.
max-txn-ttl
- The longest time that a single transaction can hold locks. If this time is exceeded, the locks of a transaction might be cleared by other transactions so that this transaction cannot be successfully committed.
- Default value:
3600000
- Unit: Millisecond
- The transaction that holds locks longer than this time can only be committed or rolled back. The commit might not be successful.
- For transactions executed using the
"bulk"
DML mode, the maximum TTL can exceed the limit of this configuration item. The maximum value is the greater value between this configuration item and 24 hours.
stmt-count-limit
- The maximum number of statements allowed in a single TiDB transaction.
- Default value:
5000
- If a transaction does not roll back or commit after the number of statements exceeds
stmt-count-limit
, TiDB returns thestatement count 5001 exceeds the transaction limitation, autocommit = false
error. This configuration takes effect only in the retryable optimistic transaction. If you use the pessimistic transaction or have disabled the transaction retry, the number of statements in a transaction is not limited by this configuration.
txn-entry-size-limit
New in v4.0.10 and v5.0.0
- The size limit of a single row of data in TiDB.
- Default value:
6291456
(in bytes) - The size limit of a single key-value record in a transaction. If the size limit is exceeded, TiDB returns the
entry too large
error. The maximum value of this configuration item does not exceed125829120
(120 MB). - Starting from v7.6.0, you can use the system variable
tidb_txn_entry_size_limit
to dynamically modify the value of this configuration item. - Note that TiKV has a similar limit. If the data size of a single write request exceeds
raft-entry-max-size
, which is 8 MB by default, TiKV refuses to process this request. When a table has a row of large size, you need to modify both configurations at the same time. - The default value of
max_allowed_packet
(the maximum size of a packet for the MySQL protocol) is 67108864 (64 MiB). If a row is larger thanmax_allowed_packet
, the row gets truncated. - The default value of
txn-total-size-limit
(the size limit of a single transaction in TiDB) is 100 MiB. If you increase thetxn-entry-size-limit
value to be over 100 MiB, you need to increase thetxn-total-size-limit
value accordingly.
txn-total-size-limit
- The size limit of a single transaction in TiDB.
- Default value:
104857600
(in bytes) - In a single transaction, the total size of key-value records cannot exceed this value. The maximum value of this parameter is
1099511627776
(1 TB). Note that if you have used the binlog to serve the downstream consumer Kafka (such as thearbiter
cluster), the value of this parameter must be no more than1073741824
(1 GB). This is because 1 GB is the upper limit of a single message size that Kafka can process. Otherwise, an error is returned if this limit is exceeded. - In TiDB v6.5.0 and later versions, this configuration is no longer recommended. The memory size of a transaction will be accumulated into the memory usage of the session, and the
tidb_mem_quota_query
variable will take effect when the session memory threshold is exceeded. To be compatible with previous versions, this configuration works as follows when you upgrade from an earlier version to TiDB v6.5.0 or later:- If this configuration is not set or is set to the default value (
104857600
), after an upgrade, the memory size of a transaction will be accumulated into the memory usage of the session, and thetidb_mem_quota_query
variable will take effect. - If this configuration is not defaulted (
104857600
), it still takes effect and its behavior on controlling the size of a single transaction remains unchanged before and after the upgrade. This means that the memory size of the transaction is not controlled by thetidb_mem_quota_query
variable.
- If this configuration is not set or is set to the default value (
- When TiDB executes transactions in the
tidb_dml_type
"bulk"
mode, transaction size is not limited by the TiDB configuration itemtxn-total-size-limit
.
tcp-keep-alive
- Determines whether to enable
keepalive
in the TCP layer. - Default value:
true
tcp-no-delay
- Determines whether to enable TCP_NODELAY at the TCP layer. After it is enabled, TiDB disables the Nagle algorithm in the TCP/IP protocol and allows sending small data packets to reduce network latency. This is suitable for latency-sensitive applications with a small transmission volume of data.
- Default value:
true
cross-join
- Default value:
true
- TiDB supports executing the
JOIN
statement without any condition (theWHERE
field) of both sides tables by default; if you set the value tofalse
, the server refuses to execute when such aJOIN
statement appears.
stats-lease
- The time interval of reloading statistics, updating the number of table rows, checking whether it is needed to perform the automatic analysis, using feedback to update statistics and loading statistics of columns.
- Default value:
3s
- At intervals of
stats-lease
time, TiDB checks the statistics for updates and updates them to the memory if updates exist. - At intervals of
20 * stats-lease
time, TiDB updates the total number of rows generated by DML and the number of modified rows to the system table. - At intervals of
stats-lease
, TiDB checks for tables and indexes that need to be automatically analyzed. - At intervals of
stats-lease
, TiDB checks for column statistics that need to be loaded to the memory. - At intervals of
200 * stats-lease
, TiDB writes the feedback cached in the memory to the system table. - At intervals of
5 * stats-lease
, TiDB reads the feedback in the system table, and updates the statistics cached in the memory.
- At intervals of
- When
stats-lease
is set to 0s, TiDB periodically reads the feedback in the system table, and updates the statistics cached in the memory every three seconds. But TiDB no longer automatically modifies the following statistics-related system tables:mysql.stats_meta
: TiDB no longer automatically records the number of table rows that are modified by the transaction and updates it to this system table.mysql.stats_histograms
/mysql.stats_buckets
andmysql.stats_top_n
: TiDB no longer automatically analyzes and proactively updates statistics.mysql.stats_feedback
: TiDB no longer updates the statistics of the tables and indexes according to a part of statistics returned by the queried data.
pseudo-estimate-ratio
- The ratio of (number of modified rows)/(total number of rows) in a table. If the value is exceeded, the system assumes that the statistics have expired and the pseudo statistics will be used.
- Default value:
0.8
- The minimum value is
0
and the maximum value is1
.
force-priority
- Sets the priority for all statements.
- Default value:
NO_PRIORITY
- Value options: The default value
NO_PRIORITY
means that the priority for statements is not forced to change. Other options areLOW_PRIORITY
,DELAYED
, andHIGH_PRIORITY
in ascending order. - Since v6.1.0, the priority for all statements is determined by the TiDB configuration item
instance.tidb_force_priority
or the system variabletidb_force_priority
.force-priority
still takes effect. But ifforce-priority
andinstance.tidb_force_priority
are set at the same time, the latter takes effect.
distinct-agg-push-down
- Determines whether the optimizer executes the operation that pushes down the aggregation function with
Distinct
(such asselect count(distinct a) from t
) to Coprocessors. - Default:
false
- This variable is the initial value of the system variable
tidb_opt_distinct_agg_push_down
.
enforce-mpp
- Determines whether to ignore the optimizer's cost estimation and to forcibly use TiFlash's MPP mode for query execution.
- Default value:
false
- This configuration item controls the initial value of
tidb_enforce_mpp
. For example, when this configuration item is set totrue
, the default value oftidb_enforce_mpp
isON
.
enable-stats-cache-mem-quota
New in v6.1.0
- Controls whether to enable the memory quota for the statistics cache.
- Default value:
true
stats-load-concurrency
New in v5.4.0
- The maximum number of columns that the TiDB synchronously loading statistics feature can process concurrently.
- Default value:
0
. Before v8.2.0, the default value is5
. - Currently, the valid value range is
[0, 128]
. The value0
means the automatic mode, which automatically adjusts concurrency based on the configuration of the server. Before v8.2.0, the minimum value is1
.
stats-load-queue-size
New in v5.4.0
- The maximum number of column requests that the TiDB synchronously loading statistics feature can cache.
- Default value:
1000
- Currently, the valid value range is
[1, 100000]
.
concurrently-init-stats
New in v8.1.0 and v7.5.2
- Controls whether to initialize statistics concurrently during TiDB startup.
- Default value:
false
lite-init-stats
New in v7.1.0
- Controls whether to use lightweight statistics initialization during TiDB startup.
- Default value:
false
for versions earlier than v7.2.0,true
for v7.2.0 and later versions. - When the value of
lite-init-stats
istrue
, statistics initialization does not load any histogram, TopN, or Count-Min Sketch of indexes or columns into memory. When the value oflite-init-stats
isfalse
, statistics initialization loads histograms, TopN, and Count-Min Sketch of indexes and primary keys into memory but does not load any histogram, TopN, or Count-Min Sketch of non-primary key columns into memory. When the optimizer needs the histogram, TopN, and Count-Min Sketch of a specific index or column, the necessary statistics are loaded into memory synchronously or asynchronously (controlled bytidb_stats_load_sync_wait
). - Setting
lite-init-stats
totrue
speeds up statistics initialization and reduces TiDB memory usage by avoiding unnecessary statistics loading. For details, see Load statistics.
force-init-stats
New in v6.5.7 and v7.1.0
- Controls whether to wait for statistics initialization to finish before providing services during TiDB startup.
- Default value:
false
for versions earlier than v7.2.0,true
for v7.2.0 and later versions. - When the value of
force-init-stats
istrue
, TiDB needs to wait until statistics initialization is finished before providing services upon startup. Note that if there are a large number of tables and partitions and the value oflite-init-stats
isfalse
, settingforce-init-stats
totrue
might prolong the time it takes for TiDB to start providing services. - When the value of
force-init-stats
isfalse
, TiDB can still provide services before statistics initialization is finished, but the optimizer uses pseudo statistics to make decisions, which might result in suboptimal execution plans.
opentracing
Configuration items related to opentracing.
enable
- Enables opentracing to trace the call overhead of some TiDB components. Note that enabling opentracing causes some performance loss.
- Default value:
false
rpc-metrics
- Enables RPC metrics.
- Default value:
false
opentracing.sampler
Configuration items related to opentracing.sampler.
type
- Specifies the type of the opentracing sampler. The string value is case-insensitive.
- Default value:
"const"
- Value options:
"const"
,"probabilistic"
,"ratelimiting"
,"remote"
param
- The parameter of the opentracing sampler.
- For the
const
type, the value can be0
or1
, which indicates whether to enable theconst
sampler. - For the
probabilistic
type, the parameter specifies the sampling probability, which can be a float number between0
and1
. - For the
ratelimiting
type, the parameter specifies the number of spans sampled per second. - For the
remote
type, the parameter specifies the sampling probability, which can be a float number between0
and1
.
- For the
- Default value:
1.0
sampling-server-url
- The HTTP URL of the jaeger-agent sampling server.
- Default value:
""
max-operations
- The maximum number of operations that the sampler can trace. If an operation is not traced, the default probabilistic sampler is used.
- Default value:
0
sampling-refresh-interval
- Controls the frequency of polling the jaeger-agent sampling policy.
- Default value:
0
opentracing.reporter
Configuration items related to opentracing.reporter.
queue-size
- The queue size with which the reporter records spans in memory.
- Default value:
0
buffer-flush-interval
- The interval at which the reporter flushes the spans in memory to the storage.
- Default value:
0
log-spans
- Determines whether to print the log for all submitted spans.
- Default value:
false
local-agent-host-port
- The address at which the reporter sends spans to the jaeger-agent.
- Default value:
""
tikv-client
grpc-connection-count
- The maximum number of connections established with each TiKV.
- Default value:
4
grpc-keepalive-time
- The
keepalive
time interval of the RPC connection between TiDB and TiKV nodes. If there is no network packet within the specified time interval, the gRPC client executesping
command to TiKV to see if it is alive. - Default:
10
- Unit: second
grpc-keepalive-timeout
- The timeout of the RPC
keepalive
check between TiDB and TiKV nodes. - Default value:
3
- Unit: second
grpc-compression-type
- Specifies the compression type used for data transfer from TiDB nodes to TiKV nodes. The default value is
"none"
, which means no compression. To enable the gzip compression, set this value to"gzip"
. - Default value:
"none"
- Value options:
"none"
,"gzip"
commit-timeout
- The maximum timeout when executing a transaction commit.
- Default value:
41s
- It is required to set this value larger than twice of the Raft election timeout.
max-batch-size
- The maximum number of RPC packets sent in batch. If the value is not
0
, theBatchCommands
API is used to send requests to TiKV, and the RPC latency can be reduced in the case of high concurrency. It is recommended that you do not modify this value. - Default value:
128
max-batch-wait-time
- Waits for
max-batch-wait-time
to encapsulate the data packets into a large packet in batch and send it to the TiKV node. It is valid only when the value oftikv-client.max-batch-size
is greater than0
. It is recommended not to modify this value. - Default value:
0
- Unit: nanoseconds
batch-wait-size
- The maximum number of packets sent to TiKV in batch. It is recommended not to modify this value.
- Default value:
8
- If the value is
0
, this feature is disabled.
overload-threshold
- The threshold of the TiKV load. If the TiKV load exceeds this threshold, more
batch
packets are collected to relieve the pressure of TiKV. It is valid only when the value oftikv-client.max-batch-size
is greater than0
. It is recommended not to modify this value. - Default value:
200
copr-req-timeout
New in v7.5.0
- The timeout of a single Coprocessor request.
- Default value:
60
- Unit: second
enable-replica-selector-v2
New in v8.0.0
- Whether to use the new version of the Region replica selector when sending RPC requests to TiKV.
- Default value:
true
tikv-client.copr-cache New in v4.0.0
This section introduces configuration items related to the Coprocessor Cache feature.
capacity-mb
- The total size of the cached data. When the cache space is full, old cache entries are evicted. When the value is
0.0
, the Coprocessor Cache feature is disabled. - Default value:
1000.0
- Unit: MB
- Type: Float
txn-local-latches
Configuration items related to the transaction latch. These configuration items might be deprecated in the future. It is not recommended to use them.
enabled
- Determines whether to enable the memory lock of transactions.
- Default value:
false
capacity
- The number of slots corresponding to Hash, which automatically adjusts upward to an exponential multiple of 2. Each slot occupies 32 Bytes of memory. If set too small, it might result in slower running speed and poor performance in the scenario where data writing covers a relatively large range (such as importing data).
- Default value:
2048000
binlog
Configurations related to TiDB Binlog.
enable
- Enables or disables binlog.
- Default value:
false
write-timeout
- The timeout of writing binlog into Pump. It is not recommended to modify this value.
- Default:
15s
- unit: second
ignore-error
- Determines whether to ignore errors occurred in the process of writing binlog into Pump. It is not recommended to modify this value.
- Default value:
false
- When the value is set to
true
and an error occurs, TiDB stops writing binlog and add1
to the count of thetidb_server_critical_error_total
monitoring item. When the value is set tofalse
, the binlog writing fails and the entire TiDB service is stopped.
binlog-socket
- The network address to which binlog is exported.
- Default value: ""
strategy
- The strategy of Pump selection when binlog is exported. Currently, only the
hash
andrange
methods are supported. - Default value:
range
status
Configuration related to the status of TiDB service.
report-status
- Enables or disables the HTTP API service.
- Default value:
true
record-db-qps
- Determines whether to transmit the database-related QPS metrics to Prometheus.
- Default value:
false
record-db-label
- Determines whether to transmit the database-related QPS metrics to Prometheus.
- Supports more metircs types than
record-db-qps
, for example, duration and statements. - Default value:
false
pessimistic-txn
For pessimistic transaction usage, refer to TiDB Pessimistic Transaction Mode.
max-retry-count
- The maximum number of retries of each statement in pessimistic transactions. If the number of retries exceeds this limit, an error occurs.
- Default value:
256
deadlock-history-capacity
- The maximum number of deadlock events that can be recorded in the
INFORMATION_SCHEMA.DEADLOCKS
table of a single TiDB server. If this table is in full volume and an additional deadlock event occurs, the earliest record in the table will be removed to make place for the newest error. - Default value:
10
- Minimum value:
0
- Maximum value:
10000
deadlock-history-collect-retryable
- Controls whether the
INFORMATION_SCHEMA.DEADLOCKS
table collects the information of retryable deadlock errors. For the description of retryable deadlock errors, see Retryable deadlock errors. - Default value:
false
pessimistic-auto-commit New in v6.0.0
- Determines the transaction mode that the auto-commit transaction uses when the pessimistic transaction mode is globally enabled (
tidb_txn_mode='pessimistic'
). By default, even if the pessimistic transaction mode is globally enabled, the auto-commit transaction still uses the optimistic transaction mode. After enablingpessimistic-auto-commit
(set totrue
), the auto-commit transaction also uses pessimistic mode, which is consistent with the other explicitly committed pessimistic transactions. - For scenarios with conflicts, after enabling this configuration, TiDB includes auto-commit transactions into the global lock-waiting management, which avoids deadlocks and mitigates the latency spike brought by deadlock-causing conflicts.
- For scenarios with no conflicts, if there are many auto-commit transactions (the specific number is determined by the real scenarios. For example, the number of auto-commit transactions accounts for more than half of the total number of applications), and a single transaction operates a large data volume, enabling this configuration causes performance regression. For example, the auto-commit
INSERT INTO SELECT
statement. - When the session-level system variable
tidb_dml_type
is set to"bulk"
, the effect of this configuration in the session is equivalent to setting it tofalse
. - Default value:
false
constraint-check-in-place-pessimistic New in v6.4.0
- Controls the default value of the system variable
tidb_constraint_check_in_place_pessimistic
. - Default value:
true
isolation-read
Configuration items related to read isolation.
engines
- Controls from which engine TiDB allows to read data.
- Default value: ["tikv", "tiflash", "tidb"], indicating that the engine is automatically selected by the optimizer.
- Value options: Any combinations of "tikv", "tiflash", and "tidb", for example, ["tikv", "tidb"] or ["tiflash", "tidb"]
instance
tidb_enable_collect_execution_info
- This configuration controls whether to record the execution information of each operator in the slow query log and whether to record the usage statistics of indexes.
- Default value:
true
- Before v6.1.0, this configuration is set by
enable-collect-execution-info
.
tidb_enable_slow_log
- This configuration is used to control whether to enable the slow log feature.
- Default value:
true
- Value options:
true
orfalse
- Before v6.1.0, this configuration is set by
enable-slow-log
.
tidb_slow_log_threshold
- Outputs the threshold value of the time consumed by the slow log.
- Default value:
300
- Range:
[-1, 9223372036854775807]
- Unit: Milliseconds
- When the time consumed by a query is larger than this value, this query is considered as a slow query and its log is output to the slow query log. Note that when the output level of
log.level
is"debug"
, all queries are recorded in the slow query log, regardless of the setting of this parameter. - Before v6.1.0, this configuration is set by
slow-threshold
.
in-mem-slow-query-topn-num
New in v7.3.0
- The configuration controls the number of slowest queries that are cached in memory.
- Default value: 30
in-mem-slow-query-recent-num
New in v7.3.0
- The configuration controls the number of recently used slow queries that are cached in memory.
- Default value: 500
tidb_expensive_query_time_threshold
- This configuration is used to set the threshold value that determines whether to print expensive query logs. The difference between expensive query logs and slow query logs is:
- Slow logs are printed after the statement is executed.
- Expensive query logs print the statements that are being executed, with execution time exceeding the threshold value, and their related information.
- Default value:
60
- Range:
[10, 2147483647]
- Unit: Seconds
- Before v5.4.0, this configuration is set by
expensive-threshold
.
tidb_record_plan_in_slow_log
- This configuration is used to control whether to include the execution plan of slow queries in the slow log.
- Default value:
1
- Value options:
1
(enabled, default) or0
(disabled). - The value of this configuration will initialize the value of system variable
tidb_record_plan_in_slow_log
- Before v6.1.0, this configuration is set by
record-plan-in-slow-log
.
tidb_force_priority
- This configuration is used to change the default priority for statements executed on a TiDB server.
- Default value:
NO_PRIORITY
- The default value
NO_PRIORITY
means that the priority for statements is not forced to change. Other options areLOW_PRIORITY
,DELAYED
, andHIGH_PRIORITY
in ascending order. - Before v6.1.0, this configuration is set by
force-priority
.
max_connections
- The maximum number of connections permitted for a single TiDB instance. It can be used for resources control.
- Default value:
0
- Range:
[0, 100000]
- The default value
0
means no limit. When the value of this variable is larger than0
, and the number of connections reaches the value, the TiDB server will reject new connections from clients. - The value of this configuration will initialize the value of system variable
max_connections
- Before v6.2.0, this configuration is set by
max-server-connections
.
tidb_enable_ddl
- This configuration controls whether the corresponding TiDB instance can become a DDL owner or not.
- Default value:
true
- Possible values:
OFF
,ON
- The value of this configuration will initialize the value of the system variable
tidb_enable_ddl
- Before v6.3.0, this configuration is set by
run-ddl
.
tidb_stmt_summary_enable_persistent
New in v6.6.0
- Controls whether to enable statements summary persistence.
- Default value:
false
- For more details, see Persist statements summary.
tidb_stmt_summary_filename
New in v6.6.0
- When statements summary persistence is enabled, this configuration specifies the file to which persistent data is written.
- Default value:
tidb-statements.log
tidb_stmt_summary_file_max_days
New in v6.6.0
- When statements summary persistence is enabled, this configuration specifies the maximum number of days to keep persistent data files.
- Default value:
3
- Unit: day
- You can adjust the value based on the data retention requirements and disk space usage.
tidb_stmt_summary_file_max_size
New in v6.6.0
- When statements summary persistence is enabled, this configuration specifies the maximum size of a persistent data file.
- Default value:
64
- Unit: MiB
- You can adjust the value based on the data retention requirements and disk space usage.
tidb_stmt_summary_file_max_backups
New in v6.6.0
- When statements summary persistence is enabled, this configuration specifies the maximum number of data files that can be persisted.
0
means no limit on the number of files. - Default value:
0
- You can adjust the value based on the data retention requirements and disk space usage.
proxy-protocol
Configuration items related to the PROXY protocol.
networks
- The list of proxy server's IP addresses allowed to connect to TiDB using the PROXY protocol
- Default value: ""
- In general cases, when you access TiDB behind a reverse proxy, TiDB takes the IP address of the reverse proxy server as the IP address of the client. By enabling the PROXY protocol, reverse proxies that support this protocol, such as HAProxy, can pass the real client IP address to TiDB.
- After configuring this parameter, TiDB allows the configured source IP address to connect to TiDB using the PROXY protocol; if a protocol other than PROXY is used, this connection will be denied. If this parameter is left empty, no IP address can connect to TiDB using the PROXY protocol. The value can be an IP address (192.168.1.50) or CIDR (192.168.1.0/24) with
,
as the separator.*
means any IP addresses.
fallbackable
New in v6.5.1
- Controls whether to enable the PROXY protocol fallback mode. If this configuration item is set to
true
, TiDB can accept clients that belong toproxy-protocol.networks
to connect to TiDB without using the PROXY protocol specification or without sending the PROXY protocol header. By default, TiDB only accepts client connections that belong toproxy-protocol.networks
and send a PROXY protocol header. - Default value:
false
experimental
The experimental
section, introduced in v3.1.0, describes the configurations related to the experimental features of TiDB.
allow-expression-index
New in v4.0.0
- Controls whether an expression index can be created. Since TiDB v5.2.0, if the function in an expression is safe, you can create an expression index directly based on this function without enabling this configuration. If you want to create an expression index based on other functions, you can enable this configuration, but correctness issues might exist. By querying the
tidb_allow_function_for_expression_index
variable, you can get the functions that are safe to be directly used for creating an expression. - Default value:
false