Updated at July 17, 2026 02:08 PM

Configuration reference

This topic describes all configuration parameters provided by Tarantool.

Most of the configuration options described in this reference can be applied to a specific instance, replica set, group, or to all instances globally. To do so, you need to define the required option at the specified level.

app

Using Tarantool as an application server, you can run your own Lua applications. In the app section, you can load the application and provide an application configuration in the app.cfg section.

app.cfg

A configuration of the application loaded using app.file or app.module.

Example

In the example below, the application is loaded from the myapp.lua file placed next to the YAML configuration file:


app:  file: 'myapp.lua'  cfg:    greeting: 'Hello'

Example on GitHub: application

Type: map

Default: nil

Environment variable: TT_APP_CFG

app.file

A path to a Lua file to load an application from.

Type: string

Default: nil

Environment variable: TT_APP_FILE

app.module

A Lua module to load an application from.

Example

The app section can be placed in any configuration scope. As an example use case, you can provide different applications for storages and routers in a sharded cluster:


groups:  storages:    app:      module: storage      # ...  routers:    app:      module: router      # ...

Type: string

Default: nil

Environment variable: TT_APP_MODULE

audit_log

The audit_log section defines configuration parameters related to audit logging.

audit_log.extract_key

If set to true, the audit subsystem extracts and prints only the primary key instead of full tuples in DML events (space_insert, space_replace, space_delete). Otherwise, full tuples are logged. The option may be useful in case tuples are big.

Type: boolean

Default: false

Environment variable: TT_AUDIT_LOG_EXTRACT_KEY

audit_log.file

Specify a file for the audit log destination. You can set the file type using the audit_log.to option. If you write logs to a file, Tarantool reopens the audit log at SIGHUP.

Type: string

Default: 'var/log/{{ instance_name }}/audit.log'

Environment variable: TT_AUDIT_LOG_FILE

audit_log.filter

Enable logging for a specified subset of audit events. This option accepts the following values:

Event names (for example, password_change). For details, see Audit log events.
Event groups (for example, audit). For details, see Event groups.

The option contains either one value from Possible values section (see below) or a combination of them.

To enable custom audit log events, specify the custom value in this option.

Example


filter: [ user_create,data_operations,ddl,custom ]  format: json  spaces: [ bands ]  extract_key: truegroups:  group001:    replicasets:      replicaset001:        instances:          instance001:            iproto:              listen:              - uri: '127.0.0.1:3301'

Type: array

Possible values: 'all', 'audit', 'auth', 'priv', 'ddl', 'dml', 'data_operations', 'compatibility', 'audit_enable', 'auth_ok', 'auth_fail', 'disconnect', 'user_create', 'user_drop', 'role_create', 'role_drop', 'user_disable', 'user_enable', 'user_grant_rights', 'role_grant_rights', 'role_revoke_rights', 'password_change', 'access_denied', 'eval', 'call', 'space_select', 'space_create', 'space_alter', 'space_drop', 'space_insert', 'space_replace', 'space_delete', 'custom' Default: 'nil'

Environment variable: TT_AUDIT_LOG_FILTER

audit_log.format

Specify a format that is used for the audit log.

Example

If you set the option to plain,


audit_log:  to: file  format: plain

the output in the file might look as follows:


2024-01-17T00:12:27.155+03004b5a2624-28e5-4b08-83c7-035a0c5a1db9INFO remote:unix/:(socket)session_type:consolemodule:tarantooluser:admintype:space_createtag:description:Create space Bands

Type: string

Possible values: 'json', 'csv', 'plain'

Default: 'json'

Environment variable: TT_AUDIT_LOG_FORMAT

audit_log.nonblock

Specify the logging behavior if the system is not ready to write. If set to true, Tarantool does not block during logging if the system is non-writable and writes a message instead. Using this value may improve logging performance at the cost of losing some log messages.

Type: boolean

Default: false

Environment variable: TT_AUDIT_LOG_NONBLOCK

audit_log.pipe

Specify a pipe for the audit log destination. You can set the pipe type using the audit_log.to option. If log is a program, its pid is stored in the audit.pid field. You need to send it a signal to rotate logs.

Example

This starts the cronolog program when the server starts and sends all audit_log messages to cronolog standard input (stdin).


audit_log:  to: pipe  pipe: 'cronolog audit_tarantool.log'

Type: string

Default: box.NULL

Environment variable: TT_AUDIT_LOG_PIPE

audit_log.spaces

The array of space names for which data operation events (space_select, space_insert, space_replace, space_delete) should be logged. The array accepts string values. If set to box.NULL, the data operation events are logged for all spaces.

Example

In the example, only the events of bands and singers spaces are logged:


audit_log:  spaces: [bands, singers]

Type: array

Default: box.NULL

Environment variable: TT_AUDIT_LOG_SPACES

audit_log.to

Enable audit logging and define the log location. This option accepts the following values:

devnull: disable audit logging.
file: write audit logs to a file (see audit_log.file).
pipe: start a program and write audit logs to it (see audit_log.pipe).
syslog: write audit logs to a system logger (see audit_log.syslog.*).

By default, audit logging is disabled.

Example

The basic audit log configuration might look as follows:


audit_log:  to: file  file: 'audit_tarantool.log'  filter: [ user_create,data_operations,ddl,custom ]  format: json  spaces: [ bands ]  extract_key: true

Type: string

Possible values: 'devnull', 'file', 'pipe', 'syslog'

Default: 'devnull'

Environment variable: TT_AUDIT_LOG_TO

audit_log.syslog.*

audit_log.syslog.facility

Specify a system logger keyword that tells syslogd where to send the message. You can enable logging to a system logger using the audit_log.to option.

audit_log.syslog.identity

Specify an application name to show in logs. You can enable logging to a system logger using the audit_log.to option.

audit_log.syslog.server

Set a location for the syslog server. It can be a Unix socket path starting with unix: or an ipv4 port number. You can enable logging to a system logger using the audit_log.to option.

Example


/code_snippets/snippets/config/instances.enabled/audit_log_syslog/config.yaml

audit_log.syslog.server – a syslog server location.
audit_log.syslog.facility – a system logger keyword that tells syslogd where to send the message. The default value is local7.
audit_log.syslog.identity – an application name to show in logs. The default value is tarantool.

These options are interpreted as a message for the syslogd program, which runs in the background of any Unix-like platform.

An example of a Tarantool audit log entry in the syslog:


09:32:52 tarantool_audit: {"time": "2024-02-08T09:32:52.190+0300", "uuid": "94454e46-9a0e-493a-bb9f-d59e44a43581", "severity": "INFO", "remote": "unix/:(socket)", "session_type": "console", "module": "tarantool", "user": "admin", "type": "space_create", "tag": "", "description": "Create space bands"}

Type: string

Default: box.NULL

Environment variable: TT_AUDIT_LOG_SYSLOG_SERVER

compat

The compat section defines values of the compat module options.

compat.binary_data_decoding

Define how to store binary data fields in Lua after decoding:

new: as varbinary objects
old: as plain strings

compat.box_cfg_replication_sync_timeout

Set a default replication sync timeout:

new: 0
old: 300 seconds

Type: string

Possible values: 'new', 'old'

Default: 'new'

Environment variable: TT_COMPAT_BOX_CFG_REPLICATION_SYNC_TIMEOUT

compat.box_error_serialize_verbose

Since: 3.1.0

Set the verbosity of error objects serialization:

new: serialize the error message together with other potentially useful fields
old: serialize only the error message

Type: string

Possible values: 'new', 'old'

Default: 'old'

Environment variable: TT_COMPAT_BOX_ERROR_SERIALIZE_VERBOSE

compat.box_error_unpack_type_and_code

Since: 3.1.0

Whether to show error fields in box.error.unpack():

new: do not show base_type and custom_type fields; do not show the code field if it is 0. Note that base_type is still accessible for an error object.
old: show all fields

Type: string

Possible values: 'new', 'old'

Default: 'old'

Environment variable: TT_COMPAT_BOX_ERROR_UNPACK_TYPE_AND_CODE

compat.box_info_cluster_meaning

Define the behavior of box.info.cluster:

new: show the entire cluster
old:: show the current replica set

Type: string

Possible values: 'new', 'old'

Default: 'new'

Environment variable: TT_COMPAT_BOX_INFO_CLUSTER_MEANING

compat.box_session_push_deprecation

Whether to raise errors on attempts to call the deprecated function box.session.push:

new: raise an error
old: do not raise an error

Type: string

Possible values: 'new', 'old'

Default: 'old'

Environment variable: TT_COMPAT_BOX_SESSION_PUSH_DEPRECATION

compat.box_space_execute_priv

Whether the execute privilege can be granted on spaces:

new: an error is raised
old: the privilege can be granted with no actual effect

Type: string

Possible values: 'new', 'old'

Default: 'new'

Environment variable: TT_COMPAT_BOX_SPACE_EXECUTE_PRIV

compat.box_space_max

Set the maximum space identifier (box.schema.SPACE_MAX):

new: 2147483646
old: 2147483647

The limit was decremented because the old max value is used as an error indicator in the box C API.

Type: string

Possible values: 'new', 'old'

Default: 'new'

Environment variable: TT_COMPAT_BOX_SPACE_MAX

compat.box_tuple_extension

Controls IPROTO_FEATURE_CALL_RET_TUPLE_EXTENSION and IPROTO_FEATURE_CALL_ARG_TUPLE_EXTENSION feature bits that define tuple encoding in iproto call and eval requests.

new: tuples with formats are encoded as MP_TUPLE
old: tuples with formats are encoded as MP_ARRAY

Type: string

Possible values: 'new', 'old'

Default: 'new'

Environment variable: TT_COMPAT_BOX_TUPLE_EXTENSION

compat.box_tuple_new_vararg

Controls how box.tuple.new interprets an argument list:

new: as a value with a tuple format
old: as an array of tuple fields

Type: string

Possible values: 'new', 'old'

Default: 'new'

Environment variable: TT_COMPAT_BOX_TUPLE_NEW_VARARG

compat.c_func_iproto_multireturn

Controls wrapping of multiple results of a stored C function when returning them via iproto:

new: return without wrapping (consistently with a local call via box.func)
old: wrap results into a MessagePack array

Type: string

Possible values: 'new', 'old'

Default: 'new'

Environment variable: TT_COMPAT_C_FUNC_IPROTO_MULTIRETURN

compat.fiber_channel_close_mode

Define the behavior of fiber channels after closing:

new: mark the channel read-only
old: destroy the channel object

compat.fiber_slice_default

Define the maximum fiber execution time without a yield:

new: {warn = 0.5, err = 1.0}
old: infinity (no warnings or errors raised).

compat.json_escape_forward_slash

Whether to escape the forward slash symbol '/' using a backslash in a json.encode() result:

new: do not escape the forward slash
old: escape the forward slash

compat.sql_priv

Whether to enable access checks for SQL requests over iproto:

new: check the user's access permissions
old: allow any user to execute SQL over iproto

Type: string

Possible values: 'new', 'old'

Default: 'new'

Environment variable: TT_COMPAT_SQL_PRIV

compat.sql_seq_scan_default

Controls the default value of the sql_seq_scan session setting:

new: false
old: true

compat.yaml_pretty_multiline

Whether to encode in block scalar style all multiline strings or ones containing the \n\n substring:

new: all multiline strings
old: only strings containing the \n\n substring

conditional

The conditional section defines the configuration parts that apply to instances that meet certain conditions.

conditional.if

conditional.if

Specify a conditional section of the configuration. The configuration options defined inside a conditional.if section apply only to instances on which the specified condition is true.

Conditions can include one variable – tarantool_version: a three-number Tarantool version running on the instance, for example, 3.1.0. It compares to version literal values that include three numbers separated by periods (x.y.z).

The following operators are available in conditions:

comparison: >, <, >=, <=, ==, !=
logical operators || (OR) and && (AND)
parentheses ()

Example:

In this example, different configuration parts apply to instances running Tarantool versions above and below 3.1.0:

On versions less than 3.1.0, the upgraded label is set to false.
On versions 3.1.0 or newer, the upgraded label is set to true. Additionally, new compat options are defined. These options were introduced in version 3.1.0, so on older versions they would cause an error.


conditional:  - if: tarantool_version < 3.1.0    labels:      upgraded: 'false'  - if: tarantool_version >= 3.1.0    labels:      upgraded: 'true'    compat:      box_error_serialize_verbose: 'new'      box_error_unpack_type_and_code: 'new'

config

The config section defines various parameters related to centralized configuration.

config.reload

Specify how the configuration is reloaded. This option accepts the following values:

auto: configuration is reloaded automatically when it is changed.
manual: configuration should be reloaded manually. In this case, you can reload the configuration in the application code using config:reload().

config.context.*

This section describes options related to loading configuration settings from external storage such as external files or environment variables.

config.context
- config.context.

config.context

Specify how to load settings from external storage. For example, this option can be used to load passwords from safe storage. You can find examples in the configuration_credentials_loading_secrets section.

Type: map

Default: nil

Environment variable: TT_CONFIG_CONTEXT

config.context.

The name of an entity that identifies a configuration value to load.

config.context..envid=config_context_name_env]}

The name of an environment variable to load a configuration value from. To load a configuration value from an environment variable, set config.context..env to env.

Example

In this example, passwords are loaded from the DBADMIN_PASSWORD and SAMPLEUSER_PASSWORD environment variables:


config:  context:    dbadmin_password:      from: env      env: DBADMIN_PASSWORD    sampleuser_password:      from: env      env: SAMPLEUSER_PASSWORD

See also: [configuration_credentials_loading_secrets](../../platform/connections_and_auth/credentials#configuration_credentials_loading_secrets {params[]}

)

config.context..rstrip

Default: nil

Environment variable: TT_CONFIG_ETCD_SSL_CA_FILE

config.etcd.ssl.ca_path

A path to a directory holding certificates to verify the peer with.

Type: string

Default: nil

Environment variable: TT_CONFIG_ETCD_SSL_CA_PATH

config.etcd.ssl.ssl_cert

Since: 3.2.0

A path to an SSL certificate file.

Type: string

Default: nil

Environment variable: TT_CONFIG_ETCD_SSL_SSL_CERT

config.etcd.ssl.ssl_key

A path to a private SSL key file.

Type: string

Default: nil

Environment variable: TT_CONFIG_ETCD_SSL_SSL_KEY

config.etcd.ssl.verify_host

Enable verification of the certificate's name (CN) against the specified host.

Type: boolean

Default: nil

Environment variable: TT_CONFIG_ETCD_SSL_VERIFY_HOST

config.etcd.ssl.verify_peer

Enable verification of the peer's SSL certificate.

Type: boolean

Default: nil

Environment variable: TT_CONFIG_ETCD_SSL_VERIFY_PEER

config.etcd.http.request.timeout

A time period required to process an HTTP request to an etcd server: from sending a request to receiving a response.

See also: etcd_local_configuration

Type: number

Default: nil

Environment variable: TT_CONFIG_ETCD_HTTP_REQUEST_TIMEOUT

This section describes options related to providing connection settings to a centralized Tarantool-based storage.

config.storage.endpoints

An array of endpoints used to access a configuration storage. Each endpoint can include the following fields:

uri: a URI of the configuration storage's instance.
login: a username used to connect to the instance.
password: a password used for authentication.
params: SSL parameters required for encrypted connections ((uri.params.* )).

Type: array

Default: nil

Environment variable: TT_CONFIG_STORAGE_ENDPOINTS

config.storage.prefix

A key prefix used to search a configuration in a centralized configuration storage. Tarantool searches keys by the following path: <prefix>/config/*. Note that <prefix> should start with a slash (/).

Type: string

Default: nil

Environment variable: TT_CONFIG_STORAGE_PREFIX

config.storage.reconnect_after

A number of seconds to wait before reconnecting to a configuration storage.

Type: number

Default: 3

Environment variable: TT_CONFIG_STORAGE_RECONNECT_AFTER

config.storage.timeout

The interval (in seconds) to perform the status check of a configuration storage.

Type: number

Default: 3

Environment variable: TT_CONFIG_STORAGE_TIMEOUT

console

Configure the administrative console. A client to the console is tt connect.

console.enabled

Whether to listen on the Unix socket provided in the console.socket option.

If the option is set to false, the administrative console is disabled.

Type: boolean

Default: true

Environment variable: TT_CONSOLE_ENABLED

console.socket

The Unix socket for the administrative console.

Mind the following nuances:

Only a Unix domain socket is allowed. A TCP socket can't be configured this way.
console.socket is a file path, without any unix: or unix/: prefixes.
If the file path is a relative path, it is interpreted relative to process.work_dir.

Type: string

Default: 'var/run/{{ instance_name }}/tarantool.control'

Environment variable: TT_CONSOLE_SOCKET

credentials

The credentials section allows you to create users and grant them the specified privileges. Learn more in configuration_credentials.

credentials.roles.*
credentials.users.*
(user_or_role_name.privileges.* )

credentials.roles.*

credentials.roles

An array of roles that can be granted to users or other roles.

Example

In the example below, the writers_space_reader role gets privileges to select data in the writers space:


roles:  writers_space_reader:    privileges:    - permissions: [ read ]      spaces: [ writers ]

Type: map

Default: nil

Environment variable: TT_CREDENTIALS_ROLES

credentials.roles.<role_name>.roles

An array of roles granted to this role.

credentials.roles.<role_name>.privileges

An array of privileges granted to this role.

See <user_or_role_name>.privileges.*.

credentials.users.*

credentials.users

credentials.users

An array of users.

Example

In this example, sampleuser gets the following privileges:

Privileges granted to the writers_space_reader role.
Privileges to select and modify data in the books space.


sampleuser:  password: '123456'  roles: [ writers_space_reader ]  privileges:  - permissions: [ read, write ]    spaces: [ books ]

Type: map

Default: nil

Environment variable: TT_CREDENTIALS_USERS

credentials.users..password

A user's password.

Example

In the example below, a password for the dbadmin user is set:


credentials:  users:    dbadmin:      password: 'T0p_Secret_P@$$w0rd'

credentials.users..roles

An array of roles granted to this user.

credentials.users..privileges

An array of privileges granted to this user.

See <user_or_role_name>.privileges.*.

<user_or_role_name>.privileges.*

<user_or_role_name>.privileges

<user_or_role_name>.privileges

Privileges that can be granted to a user or role using the following options:

<user_or_role_name>.privileges.permissions

Permissions assigned to this user or a user with this role.

Example

In this example, sampleuser gets privileges to select and modify data in the books space:


sampleuser:  password: '123456'  roles: [ writers_space_reader ]  privileges:  - permissions: [ read, write ]    spaces: [ books ]

<user_or_role_name>.privileges.spaces

Spaces to which this user or a user with this role gets the specified permissions.

Example

In this example, sampleuser gets privileges to select and modify data in the books space:


sampleuser:  password: '123456'  roles: [ writers_space_reader ]  privileges:  - permissions: [ read, write ]    spaces: [ books ]

<user_or_role_name>.privileges.functions

Functions to which this user or a user with this role gets the specified permissions.

<user_or_role_name>.privileges.sequences

Sequences to which this user or a user with this role gets the specified permissions.

<user_or_role_name>.privileges.lua_eval

Whether this user or a user with this role can execute arbitrary Lua code.

<user_or_role_name>.privileges.lua_call

A list of global user-defined Lua functions that this user or a user with this role can call. To allow calling a specific function, specify its name as the value. To allow calling all global Lua functions except built-in ones functions, specify the all value.

This option should be configured together with the execute permission.

Since version 3.3.0, the lua_call option allows granting users privileges to call specified lua function on the instance in runtime (thus it doesn't require an ability to write to the database).

Example to grant custom functions to the 'alice' user:


credentials:  users:    alice:      privileges:        - permissions: [execute]          lua_call: [my_func, my_func2]

<user_or_role_name>.privileges.sql

Whether this user or a user with this role can execute an arbitrary SQL expression.

database

The database section defines database-specific configuration parameters, such as an instance's read-write mode or transaction isolation level.

database.hot_standby

Whether to start the server in the hot standby mode. This mode can be used to provide failover without replication.

Suppose there are two cluster applications. Each cluster has one instance with the same configuration:


groups:  group001:    replicasets:      replicaset001:        instances:          instance001:            database:              hot_standby: true            wal:              dir: /tmp/wals            snapshot:              dir: /tmp/snapshots            iproto:              listen:              - uri: '127.0.0.1:3301'

In particular, both instances use the same directory for storing write-ahead logs and snapshots.

When you start both cluster applications on the same machine, the instance from the first one will be the primary instance and the second will be the standby instance. In the logs of the second cluster instance, you should see a notification:


main/104/interactive I> Entering hot standby mode

This means that the standby instance is ready to take over if the primary instance goes down. The standby instance initializes and tries to take a lock on a directory for storing write-ahead logs but fails because the primary instance has made a lock on this directory.

If the primary instance goes down for any reason, the lock is released. In this case, the standby instance succeeds in taking the lock and becomes the primary instance.

database.hot_standby has no effect:

If wal.mode is set to none.
If wal.dir_rescan_delay is set to a large value on macOS or FreeBSD. On these platforms, the hot standby mode is designed so that the loop repeats every wal.dir_rescan_delay seconds.
For spaces created with engine set to vinyl.

Examples on GitHub: hot_standby_1, hot_standby_2

Type: boolean

Default: false

Environment variable: TT_DATABASE_HOT_STANDBY

database.instance_uuid

An instance UUID.

By default, instance UUIDs are generated automatically. database.instance_uuid can be used to specify an instance identifier manually.

UUIDs should follow these rules:

The values must be true unique identifiers, not shared by other instances or replica sets within the common infrastructure.
The values must be used consistently, not changed after the initial setup. The initial values are stored in snapshot files and are checked whenever the system is restarted.
The values must comply with RFC 4122. The nil UUID is not allowed.

database.mode

An instance's operating mode. This option is in effect if replication.failover is set to off.

The following modes are available:

rw: an instance is in read-write mode.
ro: an instance is in read-only mode.

If not specified explicitly, the default value depends on the number of instances in a replica set. For a single instance, the rw mode is used, while for multiple instances, the ro mode is used.

Example

You can set the database.mode option to rw on all instances in a replica set to make a master-master configuration. In this case, replication.failover should be set to off.


replication:  failover: off


groups:  group001:    replicasets:      replicaset001:        instances:          instance001:            database:              mode: rw            iproto:              listen:              - uri: '127.0.0.1:3301'          instance002:            database:              mode: rw            iproto:              listen:              - uri: '127.0.0.1:3302'


credentials:  users:    replicator:      password: 'topsecret'      roles: [replication]


iproto:  advertise:    peer:      login: replicator


instance001:  database:    mode: rw  iproto:    listen:    - uri: '127.0.0.1:3301'instance002:  database:    mode: rw  iproto:    listen:    - uri: '127.0.0.1:3302'

Type: string

Default: box.NULL (the actual default value depends on the number of instances in a replica set) Environment variable: TT_DATABASE_MODE

database.replicaset_uuid

A replica set UUID.

By default, replica set UUIDs are generated automatically. database.replicaset_uuid can be used to specify a replica set identifier manually.

database.txn_isolation

A transaction isolation level.

Type: string

Default: best-effort

Possible values: best-effort`, `read-committed`, `read-confirmed

Environment variable: TT_DATABASE_TXN_ISOLATION

database.txn_timeout

A timeout (in seconds) after which the transaction is rolled back.

database.use_mvcc_engine

Whether the transactional manager is enabled.

Type: boolean

Default: false

Environment variable: TT_DATABASE_USE_MVCC_ENGINE

failover

The failover section defines parameters related to a supervised failover.

failover.log.to

Since: 3.3.0

Define a location Tarantool sends failover logs to. This option accepts the following values:

stderr: write logs to the standard error stream.
file: write logs to a file (see failover.log.file).

Type: string

Default: 'stderr'

Environment variable: TT_FAILOVER_LOG_TO

failover.log.file

Since: 3.3.0

Specify a file for failover logs destination. To write logs to a file, set failover.log.to to file. Otherwise, failover.log.file is ignored.

Example

The example below shows how to write failover logs to a file placed in the specified directory:


failover:  log:    to: file    file: var/log/failover.log

The feedback section describes configuration parameters for sending information about a running Tarantool instance to the specified feedback server.

feedback.crashinfo

Whether to send crash information in the case of an instance failure. This information includes:

General information from the uname output.
Build information.
The crash reason.
The stack trace.

To turn off sending crash information, set this option to false.

Type: boolean

Default: true

Environment variable: TT_FEEDBACK_CRASHINFO

feedback.enabled

Whether to send information about a running instance to the feedback server. To turn off sending feedback, set this option to false.

Type: boolean

Default: true

Environment variable: TT_FEEDBACK_ENABLED

feedback.host

The address to which information is sent.

Type: string

Default: <https://feedback.tarantool.io>

Environment variable: TT_FEEDBACK_HOST

feedback.interval

The interval (in seconds) of sending information.

Type: number

Default: 3600

Environment variable: TT_FEEDBACK_INTERVAL

feedback.metrics_collect_interval

The interval (in seconds) for collecting metrics.

Type: number

Default: 60

Environment variable: TT_FEEDBACK_METRICS_COLLECT_INTERVAL

feedback.metrics_limit

The maximum size of memory (in bytes) used to store metrics before sending them to the feedback server. If the size of collected metrics exceeds this value, earlier metrics are dropped.

Type: integer

Default: 1024 * 1024 (1048576)

Environment variable: TT_FEEDBACK_METRICS_LIMIT

feedback.send_metrics

Whether to send metrics to the feedback server. Note that all collected metrics are dropped after sending them to the feedback server.

Type: boolean

Default: true

Environment variable: TT_FEEDBACK_SEND_METRICS

fiber

The fiber section describes options related to configuring fibers, yields, and cooperative multitasking.

fiber.io_collect_interval

The time period (in seconds) a fiber sleeps between iterations of the event loop.

fiber.io_collect_interval can be used to reduce CPU load in deployments where the number of client connections is large, but requests are not so frequent (for example, each connection issues just a handful of requests per second).

Type: number

Default: box.NULL

Environment variable: TT_FIBER_IO_COLLECT_INTERVAL

fiber.too_long_threshold

If processing a request takes longer than the given period (in seconds), the fiber warns about it in the log.

fiber.too_long_threshold has effect only if log.level is greater than or equal to 4 (warn).

Type: number

Default: 0.5

Environment variable: TT_FIBER_TOO_LONG_THRESHOLD

fiber.worker_pool_threads

The maximum number of threads to use during execution of certain internal processes (for example, socket.getaddrinfo() and coio_call()).

Type: number

Default: 4

Environment variable: TT_FIBER_WORKER_POOL_THREADS

fiber.slice.*

This section describes options related to configuring time periods for fiber slices. See fiber.set_max_slice for details and examples.

fiber.slice.warn

Set a time period (in seconds) that specifies the warning slice.

Type: number

Default: 0.5

Environment variable: TT_FIBER_SLICE_WARN

fiber.slice.err

Set a time period (in seconds) that specifies the error slice.

Type: number

Default: 1

Environment variable: TT_FIBER_SLICE_ERR

fiber.top.*

This section describes options related to configuring the fiber.top() function, normally used for debug purposes. fiber.top() shows all alive fibers and their CPU consumption.

fiber.top.enabled

fiber.top.enabled

Enable or disable the fiber.top() function.

Enabling fiber.top() slows down fiber switching by about 15%, so it is disabled by default.

Type: boolean

Default: false

Environment variable: TT_FIBER_TOP_ENABLED

flightrec

The flightrec section describes options related to the flight recorder configuration.

flightrec.enabled

Enable the flight recorder.

Type: boolean

Default: false

Environment variable: TT_FLIGHTREC_ENABLED

flightrec.logs_size

Specify the size (in bytes) of the log storage. You can set this option to 0 to disable the log storage.

Type: integer

Default: 10485760

Environment variable: TT_FLIGHTREC_LOGS_SIZE

flightrec.logs_max_msg_size

Specify the maximum size (in bytes) of the log message. The log message is truncated if its size exceeds this limit.

Type: integer

Default: 4096

Maximum: 16384

Environment variable: TT_FLIGHTREC_LOGS_MAX_MSG_SIZE

flightrec.logs_log_level

Specify the level of detail the log has. The default value is 6 (VERBOSE). You can learn more about log levels from the log_level option description. Note that the flightrec.logs_log_level value might differ from log_level.

Type: integer

Default: 6

Environment variable: TT_FLIGHTREC_LOGS_LOG_LEVEL

flightrec.metrics_period

Specify the time period (in seconds) that defines how long metrics are stored from the moment of dump. So, this value defines how much historical metrics data is collected up to the moment of crash. The frequency of metric dumps is defined by flightrec.metrics_interval.

Type: integer

Default: 180

Environment variable: TT_FLIGHTREC_METRICS_PERIOD

flightrec.metrics_interval

Specify the time interval (in seconds) that defines the frequency of dumping metrics. This value shouldn't exceed flightrec.metrics_period.

Type: number

Default: 1.0

Minimum: 0.001

Environment variable: TT_FLIGHTREC_METRICS_INTERVAL

flightrec.requests_size

Specify the size (in bytes) of storage for the request and response data. You can set this parameter to 0 to disable a storage of requests and responses.

Type: integer

Default: 10485760

Environment variable: TT_FLIGHTREC_REQUESTS_SIZE

flightrec.requests_max_req_size

Specify the maximum size (in bytes) of a request entry. A request entry is truncated if this size is exceeded.

Type: integer

Default: 16384

Environment variable: TT_FLIGHTREC_REQUESTS_MAX_REQ_SIZE

flightrec.requests_max_res_size

Specify the maximum size (in bytes) of a response entry. A response entry is truncated if this size is exceeded.

Type: integer

Default: 16384

Environment variable: TT_FLIGHTREC_REQUESTS_MAX_RES_SIZE

iproto

The iproto section is used to configure parameters related to communicating to and between cluster instances.

iproto.listen

An array of URIs used to listen for incoming requests. If required, you can enable SSL for specific URIs by providing additional parameters ((uri.params.* )).

Note that a URI value can't contain parameters, a login, or a password.

Example

In the example below, iproto.listen is set explicitly for each instance in a cluster:


groups:  group001:    replicasets:      replicaset001:        instances:          instance001:            iproto:              listen:              - uri: '127.0.0.1:3301'          instance002:            iproto:              listen:              - uri: '127.0.0.1:3302'          instance003:            iproto:              listen:              - uri: '127.0.0.1:3303'

iproto.net_msg_max

To handle messages, Tarantool allocates fibers. To prevent fiber overhead from affecting the whole system, Tarantool restricts how many messages the fibers handle, so that some pending requests are blocked.

On powerful systems, increase net_msg_max, and the scheduler starts processing pending requests immediately.
On weaker systems, decrease net_msg_max, and the overhead may decrease. However, this may take some time because the scheduler must wait until already-running requests finish.

When net_msg_max is reached, Tarantool suspends processing of incoming packages until it has processed earlier messages. This is not a direct restriction of the number of fibers that handle network messages, rather it is a system-wide restriction of channel bandwidth. This in turn restricts the number of incoming network messages that the transaction processor thread handles, and therefore indirectly affects the fibers that handle network messages.

Type: integer

Default: 768

Environment variable: TT_IPROTO_NET_MSG_MAX

iproto.readahead

The size of the read-ahead buffer associated with a client connection. The larger the buffer, the more memory an active connection consumes, and the more requests can be read from the operating system buffer in a single system call.

The recommendation is to make sure that the buffer can contain at least a few dozen requests. Therefore, if a typical tuple in a request is large, e.g. a few kilobytes or even megabytes, the read-ahead buffer size should be increased. If batched request processing is not used, it's prudent to leave this setting at its default.

Type: integer

Default: 16320

Environment variable: TT_IPROTO_READAHEAD

iproto.threads

The number of network threads. There can be unusual workloads where the network thread is 100% loaded and the transaction processor thread is not, so the network thread is a bottleneck. In that case, set iproto_threads to 2 or more. The operating system kernel determines which connection goes to which thread.

Type: integer

Default: 1

Environment variable: TT_IPROTO_THREADS

iproto.advertise.*

iproto.advertise.client

A URI used to advertise the current instance to clients.

The iproto.advertise.client option accepts a URI in the following formats:

An address: host:port.
A Unix domain socket: unix/:.

Note that this option doesn't allow to set a username and password. If a remote client needs this information, it should be delivered outside of the cluster configuration.

Type: string

Default: [box.NULL](../reference_lua/box/box_null#box-null)

Environment variable: TT_IPROTO_ADVERTISE_CLIENT

iproto.advertise.peer

Settings used to advertise the current instance to other cluster members. The format of these settings is described in iproto.advertise..

Example

In the example below, the following configuration options are specified:

In the credentials section, the replicator user with the replication role is created.
iproto.advertise.peer specifies that other instances should connect to an address defined in iproto.listen using the replicator user.


credentials:  users:    replicator:      password: 'topsecret'      roles: [replication]iproto:  advertise:    peer:      login: replicatorreplication:  failover: electiongroups:  group001:    replicasets:      replicaset001:        instances:          instance001:            iproto:              listen:              - uri: '127.0.0.1:3301'          instance002:            iproto:              listen:              - uri: '127.0.0.1:3302'          instance003:            iproto:              listen:              - uri: '127.0.0.1:3303'

Type: map

Environment variable: see [iproto.advertise.](peer_or_sharding/.* )

iproto.advertise.sharding

Settings used to advertise the current instance to a router and rebalancer. The format of these settings is described in iproto.advertise..

Example

In the example below, the following configuration options are specified:

In the credentials section, the replicator and storage users are created.
iproto.advertise.peer specifies that other instances should connect to an address defined in iproto.listen with the replicator user.
iproto.advertise.sharding specifies that a router should connect to storages using an address defined in iproto.listen with the storage user.


credentials:  users:    replicator:      password: 'topsecret'      roles: [replication]    storage:      password: 'secret'      roles: [sharding]iproto:  advertise:    peer:      login: replicator    sharding:      login: storage

Type: map

Environment variable: see [iproto.advertise.](peer_or_sharding/.* )

iproto.advertise.<peer_or_sharding>.*

iproto.advertise.<peer_or_sharding>.uri

(Optional) A URI used to advertise the current instance. By default, the URI defined in iproto.listen is used to advertise the current instance.

Type: string

Default: nil

Environment variable: TT_IPROTO_ADVERTISE_PEER_URI, TT_IPROTO_ADVERTISE_SHARDING_URI

iproto.advertise.<peer_or_sharding>.login

(Optional) A username used to connect to the current instance. If a username is not set, the guest user is used.

Type: string

Default: nil

Environment variable: TT_IPROTO_ADVERTISE_PEER_LOGIN, TT_IPROTO_ADVERTISE_SHARDING_LOGIN

iproto.advertise.<peer_or_sharding>.password

(Optional) A password for the specified user. If a login is specified but a password is missing, it is taken from the user's credentials.

Type: string

Default: nil

Environment variable: TT_IPROTO_ADVERTISE_PEER_PASSWORD, TT_IPROTO_ADVERTISE_SHARDING_PASSWORD

iproto.advertise.<peer_or_sharding>.params

(Optional) URI parameters required for connecting to the current instance.

.params.*

URI parameters that can be used in the iproto.listen. and iproto.advertise. options.

.params.transport

Allows you to enable traffic encryption for client-server communications over binary connections. In a Tarantool cluster, one instance might act as the server that accepts connections from other instances and the client that connects to other instances.

<uri>.params.transport accepts one of the following values:

plain (default): turn off traffic encryption.
ssl: encrypt traffic by using the TLS 1.2 protocol (Enterprise Edition only).

Example

The example below demonstrates how to enable traffic encryption by using a self-signed server certificate. The following parameters are specified for each instance:

ssl_cert_file: a path to an SSL certificate file.
ssl_key_file: a path to a private SSL key file.


replicaset001:

Example on Github: ssl_without_ca

Type: string

Default: 'plain'

Environment variable: TT_IPROTO_ADVERTISE_PEER_PARAMS_TRANSPORT, TT_IPROTO_ADVERTISE_SHARDING_PARAMS_TRANSPORT

.params.ssl_ca_file

(Optional) A path to a trusted certificate authorities (CA) file. If not set, the peer won't be checked for authenticity.

Both a server and a client can use the ssl_ca_file parameter:

If it's on the server side, the server verifies the client.
If it's on the client side, the client verifies the server.
If both sides have the CA files, the server and the client verify each other.

See also: (uri.params.transport )

Type: string

Default: nil

Environment variable: TT_IPROTO_ADVERTISE_PEER_PARAMS_SSL_CA_FILE,TT_IPROTO_ADVERTISE_SHARDING_PARAMS_SSL_CA_FILE

.params.ssl_cert_file

A path to an SSL certificate file:

For a server, it's mandatory.
For a client, it's mandatory if the ssl_ca_file parameter is set for a server; otherwise, optional.

See also: (uri.params.transport )

Type: string

Default: nil

Environment variable: TT_IPROTO_ADVERTISE_PEER_PARAMS_SSL_CERT_FILE, TT_IPROTO_ADVERTISE_SHARDING_PARAMS_SSL_CERT_FILE

.params.ssl_ciphers

(Optional) A colon-separated (:) list of SSL cipher suites the connection can use. Note that the list is not validated: if a cipher suite is unknown, Tarantool ignores it, doesn't establish the connection, and writes to the log that no shared cipher was found.

The supported cipher suites are:

ECDHE-ECDSA-AES256-GCM-SHA384
ECDHE-RSA-AES256-GCM-SHA384
DHE-RSA-AES256-GCM-SHA384
ECDHE-ECDSA-CHACHA20-POLY1305
ECDHE-RSA-CHACHA20-POLY1305
DHE-RSA-CHACHA20-POLY1305
ECDHE-ECDSA-AES128-GCM-SHA256
ECDHE-RSA-AES128-GCM-SHA256
DHE-RSA-AES128-GCM-SHA256
ECDHE-ECDSA-AES256-SHA384
ECDHE-RSA-AES256-SHA384
DHE-RSA-AES256-SHA256
ECDHE-ECDSA-AES128-SHA256
ECDHE-RSA-AES128-SHA256
DHE-RSA-AES128-SHA256
ECDHE-ECDSA-AES256-SHA
ECDHE-RSA-AES256-SHA
DHE-RSA-AES256-SHA
ECDHE-ECDSA-AES128-SHA
ECDHE-RSA-AES128-SHA
DHE-RSA-AES128-SHA
AES256-GCM-SHA384
AES128-GCM-SHA256
AES256-SHA256
AES128-SHA256
AES256-SHA
AES128-SHA
GOST2012-GOST8912-GOST8912
GOST2001-GOST89-GOST89

For detailed information on SSL ciphers and their syntax, refer to OpenSSL documentation.

See also: (uri.params.transport )

Type: string

Default: nil

Environment variable: TT_IPROTO_ADVERTISE_PEER_PARAMS_SSL_CIPHERS, TT_IPROTO_ADVERTISE_SHARDING_PARAMS_SSL_CIPHERS`

.params.ssl_key_file

A path to a private SSL key file:

For a server, it's mandatory.
For a client, it's mandatory if the ssl_ca_file parameter is set for a server; otherwise, optional.

If the private key is encrypted, provide a password for it in the ssl_password or ssl_password_file parameter.

See also: (uri.params.transport )

Type: string

Default: nil

Environment variable: TT_IPROTO_ADVERTISE_PEER_PARAMS_SSL_KEY_FILE,TT_IPROTO_ADVERTISE_SHARDING_PARAMS_SSL_KEY_FILE

.params.ssl_password

(Optional) A password for an encrypted private SSL key provided using ssl_key_file. Alternatively, the password can be provided in ssl_password_file.

Tarantool applies the ssl_password and ssl_password_file parameters in the following order:

If ssl_password is provided, Tarantool tries to decrypt the private key with it.
If ssl_password is incorrect or isn't provided, Tarantool tries all passwords from ssl_password_file one by one in the order they are written.
If ssl_password and all passwords from ssl_password_file are incorrect, or none of them is provided, Tarantool treats the private key as unencrypted.

See also: (uri.params.transport )

Type: string

Default: nil

Environment variable: TT_IPROTO_ADVERTISE_PEER_PARAMS_SSL_PASSWORD, TT_IPROTO_ADVERTISE_SHARDING_PARAMS_SSL_PASSWORD

.params.ssl_password_file

(Optional) A text file with one or more passwords for encrypted private SSL keys provided using ssl_key_file (each on a separate line). Alternatively, the password can be provided in ssl_password.

See also: (uri.params.transport )

Type: string

Default: nil

Environment variable: TT_IPROTO_ADVERTISE_PEER_PARAMS_SSL_PASSWORD_FILE, TT_IPROTO_ADVERTISE_SHARDING_PARAMS_SSL_PASSWORD_FILE

groups

The groups section provides the ability to define the full topology of a Tarantool cluster.

groups.<group_name>

A group name.

The following rules are applied to group names:

The maximum number of symbols is 63.
Should start with a letter.
Can contain lowercase letters (a-z).
Can contain digits (0-9).
Can contain the following characters: -, _.

groups.<group_name>.replicasets

Replica sets that belong to this group. See replicasets.

groups.<group_name>.<config_parameter>

Any configuration parameter that can be defined in the group scope. For example, iproto and database configuration parameters defined at the group level are applied to all instances in this group.

replicasets

replicasets.<replicaset_name>

A replica set name.

Note that the rules applied to a replica set name are the same as for groups. Learn more in groups.<group_name>.

replicasets.<replicaset_name>.leader

A replica set leader. This option can be used to set a replica set leader when manual replication.failover is used.

To perform controlled failover, <replicaset_name>.leader can be temporarily removed or set to null.

Example


replication:  failover: manualgroups:  group001:    replicasets:      replicaset001:        leader: instance001        instances:          instance001:            iproto:              listen:              - uri: '127.0.0.1:3301'          instance002:            iproto:              listen:              - uri: '127.0.0.1:3302'          instance003:            iproto:              listen:              - uri: '127.0.0.1:3303'

replicasets.<replicaset_name>.bootstrap_leader

A bootstrap leader for a replica set. To specify a bootstrap leader manually, you need to set replication.bootstrap_strategy to config.

Example


groups:  group001:    replicasets:      replicaset001:        replication:          bootstrap_strategy: config        bootstrap_leader: instance001        instances:          instance001:            iproto:              listen:              - uri: '127.0.0.1:3301'          instance002:            iproto:              listen:              - uri: '127.0.0.1:3302'          instance003:            iproto:              listen:              - uri: '127.0.0.1:3303'

replicasets.<replicaset_name>.instances

Instances that belong to this replica set. See instances.

replicasets.<replicaset_name>.<config_parameter>

Any configuration parameter that can be defined in the replica set scope. For example, iproto and database configuration parameters defined at the replica set level are applied to all instances in this replica set.

instances

instances.<instance_name>

An instance name.

Note that the rules applied to an instance name are the same as for groups. Learn more in groups.<group_name>.

instances.<instance_name>.<config_parameter>

Any configuration parameter that can be defined in the instance scope. For example, iproto and database configuration parameters defined at the instance level are applied to this instance only.

isolated mode –––

Since version 3.3.0, a new isolated option is added to instance configuration.

The option takes boolean values, by default it is set to false. isolated: true moves the instance it has been applied at to the isolated mode.

The isolated mode allows the user to temporarily isolate an instance and perform maintenance activities on it.

In the isolated mode:

The instance is moved to the read-only state
iproto stops listening for new connections
iproto drops all the current connections
The instance is disconnected from all the replication upstreams
Other replicaset members exclude the isolated instance from the replication upstreams

Example

The example below shows how to isolate an instance:


/code_snippets/snippets/config/instances.enabled/isolated_mode/config.yaml

labels

The labels section allows adding custom attributes to the configuration. Attributes must be key: value pairs with string keys and values.

labels.<label_name>

labels.<label_name>

A value of the label with the specified name.

Example

The example below shows how to define labels on the replica set and instance levels:


labels:  dc: 'east'  production: 'false'

log

The log section defines configuration parameters related to logging. To handle logging in your application, use the log module.

log.to

Define a location Tarantool sends logs to. This option accepts the following values:

stderr: write logs to the standard error stream.
file: write logs to a file (see log.file).
pipe: start a program and write logs to its standard input (see log.pipe).
syslog: write logs to a system logger (see log.syslog.*).

Type: string

Default: 'stderr'

Environment variable: TT_LOG_TO

log.file

Specify a file for logs destination. To write logs to a file, you need to set log.to to file. Otherwise, log.file is ignored.

Example

The example below shows how to write logs to a file placed in the specified directory:


log:  to: file  file: var/log/{{ instance_name }}/instance.log

Example on GitHub: log_file

Type: string

Default: 'var/log/{{ instance_name }}/tarantool.log'

Environment variable: TT_LOG_FILE

log.format

Specify a format that is used for a log entry. The following formats are supported:

plain: a log entry is formatted as plain text. Example:


2024-04-09 11:00:10.369 [12089] main/104/interactive I> log level 5 (INFO)

json: a log entry is formatted as JSON and includes additional fields. Example:


{  "time": "2024-04-09T11:00:57.174+0300",  "level": "INFO",  "message": "log level 5 (INFO)",  "pid": 12160,  "cord_name": "main",  "fiber_id": 104,  "fiber_name": "interactive",  "file": "src/main.cc",  "line": 498}

Type: string

Default: 'plain'

Environment variable: TT_LOG_FORMAT

log.level

Specify the level of detail logs have. There are the following levels:

0 – fatal
1 – syserror
2 – error
3 – crit
4 – warn
5 – info
6 – verbose
7 – debug

By setting log.level, you can enable logging of all events with severities above or equal to the given level.

Example

The example below shows how to log all events with severities above or equal to the VERBOSE level.


log:  level: 'verbose'

Example on GitHub: log_level

Type: number, string

Default: 5

Environment variable: TT_LOG_LEVEL

log.modules

Configure the specified log levels (log.level) for different modules.

You can specify a logging level for the following module types:

Modules (files) that use the default logger. Example: Set log levels for files that use the default logger.
Modules that use custom loggers created using the log.new() function. Example: Set log levels for modules that use custom loggers.
The tarantool module that enables you to configure the logging level for Tarantool core messages. Specifically, it configures the logging level for messages logged from non-Lua code, including C modules. Example: Set a log level for C modules.

Example 1: Set log levels for files that use the default logger

Suppose you have two identical modules placed by the following paths: test/module1.lua and test/module2.lua. These modules use the default logger and look as follows:


return {    say_hello = function()        local log = require('log')        log.info('Info message from module1')    end}

To configure logging levels, you need to provide module names corresponding to paths to these modules:


log:  modules:    test.module1: 'verbose'    test.module2: 'error'app:  file: 'app.lua'

To load these modules in your application (app.lua), you need to add the corresponding require directives:

Given that module1 has the verbose logging level and module2 has the error level, calling module1.say_hello() shows a message but module2.say_hello() is swallowed:

Example on GitHub: log_existing_modules

Example 2: Set log levels for modules that use custom loggers

This example shows how to set the verbose level for module1 and the error level for module2:


log:  modules:    module1: 'verbose'    module2: 'error'app:  file: 'app.lua'

To create custom loggers in your application (app.lua), call the log.new() function:

Given that module1 has the verbose logging level and module2 has the error level, calling module1_log.info() shows a message but module2_log.info() is swallowed:

Example on GitHub: log_new_modules

Example 3: Set a log level for C modules

This example shows how to set the info level for the tarantool module:


log:  modules:    tarantool: 'info'app:  file: 'app.lua'

The specified level affects messages logged from C modules:


ffi = require('ffi')-- Prints 'info' messages --ffi.C._say(ffi.C.S_INFO, nil, 0, nil, 'Info message from C module')--[[[6024] main/103/interactive I> Info message from C module---...--]]-- Swallows 'debug' messages --ffi.C._say(ffi.C.S_DEBUG, nil, 0, nil, 'Debug message from C module')--[[---...--]]

The example above uses the LuaJIT ffi library to call C functions provided by the say module.

Example on GitHub: log_existing_c_modules

Type: map

Default: [box.NULL](../reference_lua/box/box_null#box-null)

Environment variable: TT_LOG_MODULES

log.nonblock

Type: boolean

Default: false

Environment variable: TT_LOG_NONBLOCK

log.pipe

Start a program and write logs to its standard input (stdin). To send logs to a program's standard input, you need to set log.to to pipe.

Example

In the example below, Tarantool writes logs to the standard input of cronolog:


log:  to: pipe  pipe: 'cronolog tarantool.log'

Example on GitHub: log_pipe

Set a location of a syslog server. This option accepts one of the following values:

An IPv4 address. Example: 127.0.0.1:514.
A Unix socket path starting with unix:. Examples: unix:/dev/log on Linux or unix:/var/run/syslog on macOS.

To write logs to syslog, you need to set log.to to syslog.

Example

In the example below, Tarantool writes logs to a syslog server that listens for logging messages on the 127.0.0.1:514 address:


log:  to: syslog  syslog:    server: '127.0.0.1:514'

Example on GitHub: log_syslog

Type: string

Default: box.NULL

Environment variable: TT_LOG_SYSLOG_SERVER

lua

The lua section outlines the configuration parameters related to the Lua environment within Tarantool.

lua.memory

lua.memory

Specifies the maximum memory amount available to Lua scripts, measured in bytes.

When the specified value exceeds the current memory usage, the new limit takes effect immediately without a restart. However, when the specified value is lower than the current memory usage, a restart of the instance is required for the change to take effect.

Example to set the Lua memory limit to 4 GB:


lua:    memory: 4294967296

Type: integer

Default: 2147483648 (2GB)

Environment variable: TT_LUA_MEMORY

memtx

The memtx section is used to configure parameters related to the memtx engine.

memtx.allocator

Specify the allocator that manages memory for memtx tuples. Possible values:

system – the memory is allocated as needed, checking that the quota is not exceeded. THe allocator is based on the malloc function.
small – a slab allocator. The allocator repeatedly uses a memory block to allocate objects of the same type. Note that this allocator is prone to unresolvable fragmentation on specific workloads, so you can switch to system in such cases.

Type: string

Default: 'small'

Environment variable: TT_MEMTX_ALLOCATOR

memtx.max_tuple_size

Size of the largest allocation unit for the memtx storage engine in bytes. It can be increased if it is necessary to store large tuples.

Type: integer

Default: 1048576

Environment variable: TT_MEMTX_MAX_TUPLE_SIZE

memtx.memory

The amount of memory in bytes that Tarantool allocates to store tuples. When the limit is reached, INSERT and UPDATE requests fail with the ER_MEMORY_ISSUE error. The server does not go beyond the memtx.memory limit to allocate tuples, but there is additional memory used to store indexes and connection information.

Example

In the example below, the memory size is set to 1 GB (1073741824 bytes).


memtx:  memory: 1073741824

Type: integer

Default: 268435456

Environment variable: TT_MEMTX_MEMORY

memtx.min_tuple_size

Size of the smallest allocation unit in bytes. It can be decreased if most of the tuples are very small.

Type: integer

Default: 16

Possible values: between 8 and 1048280 inclusive

Environment variable: TT_MEMTX_MIN_TUPLE_SIZE

memtx.slab_alloc_factor

The multiplier for computing the sizes of memory chunks that tuples are stored in. A lower value may result in less wasted memory depending on the total amount of memory available and the distribution of item sizes.

memtx.slab_alloc_granularity

Specify the granularity in bytes of memory allocation in the small allocator. The memtx.slab_alloc_granularity value should meet the following conditions:

The value is a power of two.
The value is greater than or equal to 4.

Below are few recommendations on how to adjust the memtx.slab_alloc_granularity option:

If the tuples in space are small and have about the same size, set the option to 4 bytes to save memory.
If the tuples are different-sized, increase the option value to allocate tuples from the same mempool (memory pool).

memtx.sort_threads

The number of threads from the thread pool used to sort keys of secondary indexes on loading a memtx database. The minimum value is 1, the maximum value is 256. The default is to use all available cores.

Type: integer

Default: box.NULL

Environment variable: TT_MEMTX_SORT_THREADS

metrics

The metrics section defines configuration parameters for metrics.

metrics.exclude

An array containing the metrics to turn off. The array can contain the same values as the exclude configuration parameter passed to metrics.cfg().

Example


metrics:  include: [ all ]  exclude: [ vinyl ]  labels:    alias: '{{ instance_name }}'

Type: array

Default: ````

Environment variable: TT_METRICS_EXCLUDE

metrics.include

An array containing the metrics to turn on. The array can contain the same values as the include configuration parameter passed to metrics.cfg().

Type: array

Default: [ all ]

Environment variable: TT_METRICS_INCLUDE

metrics.labels

Global labels to be added to every observation.

Type: map

Default: { alias = names.instance_name }

Environment variable: TT_METRICS_LABELS

process

The process section defines configuration parameters of the Tarantool process in the system.

process.background

Run the server as a daemon process.

If this option is set to true, Tarantool log location defined by the log.to option should be set to file, pipe, or syslog -- anything other than stderr, the default, because a daemon process is detached from a terminal and it can't write to the terminal's stderr.

Type: boolean

Default: false

Environment variable: TT_PROCESS_BACKGROUND

process.coredump

Create coredump files.

Usually, an administrator needs to call ulimit -c unlimited (or set corresponding options in systemd's unit file) before running a Tarantool process to get core dumps. If process.coredump is enabled, Tarantool sets the corresponding resource limit by itself and the administrator doesn't need to call ulimit -c unlimited (see man 3 setrlimit).

This option also sets the state of the dumpable attribute, which is enabled by default, but may be dropped in some circumstances (according to man 2 prctl, see PR_SET_DUMPABLE).

Type: boolean

Default: false

Environment variable: TT_PROCESS_COREDUMP

process.title

Add the given string to the server's process title (it is shown in the COMMAND column for the Linux commands ps -ef and top -c).

For example, if you set the option to myservice - {{ instance_name }}:


process:  title: myservice - {{ instance_name }}

ps -ef might show the Tarantool server process like this:


$ ps -ef | grep tarantool503      68100 68098  0 10:33 pts/2    00:00.10 tarantool <running>: myservice instance1

Type: string

Default: 'tarantool - {{ instance_name }}'

Environment variable: TT_PROCESS_TITLE

process.pid_file

Store the process id in this file.

This option may contain a relative file path. In this case, it is interpreted as relative to process.work_dir.

Type: string

Default: 'var/run/{{ instance_name }}/tarantool.pid'

Environment variable: TT_PROCESS_PID_FILE

process.strip_core

Whether coredump files should not include memory allocated for tuples -- this memory can be large if Tarantool runs under heavy load. Setting to true means "do not include".

Type: boolean

Default: true

Environment variable: TT_PROCESS_STRIP_CORE

process.username

The name of the system user to switch to after start.

Type: string

Default: box.NULL

Environment variable: TT_PROCESS_USERNAME

process.work_dir

A directory where Tarantool working files will be stored (database files, logs, a PID file, a console Unix socket, and other files if an application generates them in the current directory). The server instance switches to process.work_dir with chdir(2) after start.

If set as a relative file path, it is relative to the current working directory, from where Tarantool is started. If not specified, defaults to the current working directory.

Other directory and file parameters, if set as relative paths, are interpreted as relative to process.work_dir, for example, directories for storing snapshots and write-ahead logs.

Type: string

Default: box.NULL

Environment variable: TT_PROCESS_WORK_DIR

replication

The replication section defines configuration parameters related to replication.

replication.anon

Whether to make the current instance act as an anonymous replica. Anonymous replicas are read-only and can be used, for example, for backups.

To make the specified instance act as an anonymous replica, set replication.anon to true:


instance003:  replication:    anon: true

You can find the full example on GitHub: anonymous_replica.

Anonymous replicas are not displayed in the box.info.replication section. You can check their status using box.info.replication_anon().

While anonymous replicas are read-only, you can write data to replication-local and temporary spaces (created with is_local = true and temporary = true, respectively). Given that changes to replication-local spaces are allowed, an anonymous replica might increase the 0 component of the vclock value.

Here are the limitations of having anonymous replicas in a replica set:

A replica set must contain at least one non-anonymous instance.
An anonymous replica can't be configured as a writable instance by setting database.mode to rw or making it a leader using (replicaset_name.leader ).
If replication.failover is set to election, an anonymous replica can have replication.election_mode set to off only.
If replication.failover is set to supervised, an external failover coordinator doesn't consider anonymous replicas when selecting a bootstrap or replica set leader.

Type: boolean

Default: false

Environment variable: TT_REPLICATION_ANON

replication.autoexpel

Since: 3.3.0

The replication.autoexpel option designed for managing dynamic clusters using YAML-based configurations. It enables the automatic expulsion of instances that are removed from the YAML configuration.

Only instances with names that match the specified prefix are considered for expulsion; all others are excluded. Additionally, instances without a persistent name are ignored.

If an instance is in read-write mode and has the latest database schema, it initiates the expulsion of instances that:

Match the specified prefix
Absent from the updated YAML configuration

The expulsion process follows the standard procedure, involving the removal of the instance from the _cluster system space.

The autoexpel logic is activated during specific events:

Startup. When the cluster starts, autoexpel checks and removes instances not matching the updated configuration.
Reconfiguration. When the YAML configuration is reloaded, autoexpel compares the current state to the updated configuration and performs necessary expulsions.
box.status watcher event. Changes detected by the box.status watcher also trigger the autoexpel mechanism.

autoexpel does not take any actions on newly joined instances unless one of the triggering events occurs. This means that an instance meeting the autoexpel criterion can still join the cluster, but it may be removed later during reconfiguration or on subsequent triggering events.

Configuration fields

by (string, default: nil): specifies the autoexpel criterion. Currently, only prefix is supported and must be explicitly set.
enabled (boolean, default: false): enables or disables the autoexpel logic.
prefix (string, default: nil): defines the pattern for instance names that are considered part of the cluster.

replication.autoexpel_by.* ~~~~~~~~~~~~~

replication.autoexpel_by purpose is to define the criterion used for determining which instances in a cluster are subject to the autoexpel process.

The by field helps differentiate between:

Instances that are part of the cluster and should adhere to the YAML configuration.
Instances or tools (e.g., CDC tools) that use the replication channel but are not part of the cluster configuration.

The default value of by is nil, meaning no autoexpel criterion is applied unless explicitly set.

Currently, the only supported value for by is prefix. The prefix value instructs the system to identify instances based on their names, matching them against a prefix pattern defined in the configuration.

If the autoexpel feature is enabled (enabled: true), the by field must be explicitly set to prefix.

The absence of this field or an unsupported value will result in configuration errors.


replication:  autoexpel:    enabled: true    by: prefix    prefix: '{{ replicaset_name }}'

Type: string

Default: nil

Environment variable: TT_REPLICATION_AUTOEXPEL_BY

replication.autoexpel_enabled.* ~~~~~~~~~~~~~

The replication.autoexpel_enabled field is a boolean configuration option that determines whether the autoexpel logic is active for the cluster. This feature is designed to automatically manage dynamic cluster configurations by removing instances that are no longer present in the YAML configuration.

Enabling autoexpel logic

To enable autoexpel, you should set enabled to true in the replication.autoexpel section of your YAML configuration:


replication:  autoexpel:    enabled: true    by: prefix    prefix: '{{ replicaset_name }}'

To disable autoexpel, set enabled to false.

Dependencies

If enabled is set to true, the following fields are required:

by: specifies the criterion for autoexpel (e.g., prefix).
prefix: defines the pattern used to match instance names for expulsion.

Failure to configure these fields when enabled is true will result in a configuration error.

Type: boolean

Default: false

Environment variable: TT_REPLICATION_AUTOEXPEL_ENABLED

replication.autoexpel_prefix.* ~~~~~~~~~~~~~

The prefix field filters instances for expulsion by differentiating cluster instances (from the YAML configuration) from external services (e.g., CDC tools). Only instances matching the prefix are considered.

A consistent naming pattern ensures the _cluster system space automatically aligns with the YAML configuration.

If the prefix field is not set (nil), the autoexpel logic cannot identify instances for expulsion, and the feature will not function. This field is mandatory when replication.autoexpel_enabled is set to true.

How it works:

The prefix filters instance names (e.g., {{ replicaset_name }} for replicaset-specific names or i- for names starting with i-).
Instances matching the prefix and removed from the YAML configuration are expelled.
Unnamed instances or those not matching the prefix are ignored.

Dynamic prefix based on replicaset name:


replication:  autoexpel:    enabled: true    by: prefix    prefix: '{{ replicaset_name }}'

In this setup:

Instances are grouped by replicaset names (e.g., r-001-i-001 for replicaset r-001).
The prefix ensures that only instances with names matching the replicaset name are auto expelled when removed from the configuration.

Static prefix for matching patterns:


replication:  autoexpel:    enabled: true    by: prefix    prefix: 'i-'

In this setup:

All instances with names starting with i- (e.g., i-001, i-002) are considered for expulsion.
This is useful when instances follow a uniform naming convention.

Type: string

Default: nil

Environment variable: TT_REPLICATION_AUTOEXPEL_PREFIX

autoexpel full example ~~~~~~~~~~~~~

Create a config.yaml file with the following content:


credentials:  users:    guest:      roles: [super]replication:  failover: manual  autoexpel:    enabled: true    by: prefix    prefix: '{{ replicaset_name }}'iproto:  listen:    - uri: 'unix/:./var/run/{{ instance_name }}.iproto'groups:  g-001:    replicasets:      r-001:        leader: r-001-i-001        instances:          r-001-i-001: {}          r-001-i-002: {}          r-001-i-003: {}

This configuration: - Sets up authentication with a guest user assigned the super role. - Enables the autoexpel option to automatically expel instances not present in the YAML file. - Defines instance names based on a prefix pattern: {{ replicaset_name }}. - Lists three instances: r-001-i-001, r-001-i-002, and r-001-i-003.

Open terminal window and start three instances using the following commands:


tarantool –name r-001-i-001 –config config.yaml -i


tarantool –name r-001-i-002 –config config.yaml -i


tarantool –name r-001-i-003 –config config.yaml -i

Edit config.yaml and remove the following entry for r-001-i-003:


r-001-i-003: {}

The updated config.yaml should look like this:


groups:  g-001:    replicasets:      r-001:        leader: r-001-i-001        instances:          r-001-i-001: {}          r-001-i-002: {}

Save the file.

For the leader instance (r-001-i-001), check the _cluster space:

### replication.bootstrap_strategy {params[anchor=configuration_reference_replication_bootstrap_strategy]} Specifies a strategy used to bootstrap a [replica set](../../platform/replication/replication_tutorials/repl_bootstrap#replication-bootstrap). The following strategies are available:

auto: a node doesn't boot if half or more of the other nodes in a replica set are not connected. For example, if a replica set contains 2 or 3 nodes, a node requires 2 connected instances. In the case of 4 or 5 nodes, at least 3 connected instances are required. Moreover, a bootstrap leader fails to boot unless every connected node has chosen it as a bootstrap leader.
config: use the specified node to bootstrap a replica set. To specify the bootstrap leader, use the (replicaset_name.bootstrap_leader ) option.
supervised: a bootstrap leader isn't chosen automatically but should be appointed using box.ctl.make_bootstrap_leader() on the desired node. The bootstrap leader management is the user's responsibility unless replication.failover is set to supervised. In this case, the failover coordinator manages the bootstrap leader.
native (since 3.4.0): the bootstrap leader is managed in the YAML configuration layer in sync with read-only/read-write mode management. From the user's perspective, this strategy is similar to auto: Tarantool or the failover coordinator handles bootstrapping automatically. Internally, native is based on supervised, which helps avoid limitations of auto, such as requiring a joining replica to connect to all instances registered in the _cluster system space.

When initializing a replica set, native uses box.ctl.make_bootstrap_leader({graceful . After bootstrap, when an instance switches to read-write mode, native uses box.ctl.make_bootstrap_leader() to keep the bootstrap leader record pointing to the current read-write instance. The instance chosen to initialize a replica set depends on replication.failover:
- off: the first read-write instance in lexicographic order.
- manual: the instance configured as the replica set leader.
- election: the first non-anonymous candidate in lexicographic order.
- supervised: the failover coordinator chooses the bootstrap leader.
legacy (deprecated since 2.11.0): a node requires the replication_connect_quorum number of other nodes to be connected. This option is added to keep the compatibility with the current versions of Cartridge and might be removed in the future.

Type: string

Default: auto

Environment variable: TT_REPLICATION_BOOTSTRAP_STRATEGY

replication.connect_timeout

A timeout (in seconds) a replica waits when trying to connect to a master in a cluster. See orphan status for details.

This parameter is different from replication.timeout, which a master uses to disconnect a replica when the master receives no acknowledgments of heartbeat messages.

Type: number

Default: 30

Environment variable: TT_REPLICATION_CONNECT_TIMEOUT

replication.election_mode

A role of a replica set node in the leader election process.

The possible values are:

off: a node doesn't participate in the election activities.
voter: a node can participate in the election process but can't be a leader.
candidate: a node should be able to become a leader.
manual: allow to control which instance is the leader explicitly instead of relying on automated leader election. By default, the instance acts like a voter -- it is read-only and may vote for other candidate instances. Once box.ctl.promote() is called, the instance becomes a candidate and starts a new election round. If the instance wins the elections, it becomes a leader but won't participate in any new elections.

Type: string

Default: [box.NULL](../reference_lua/box/box_null#box-null) (the actual default value depends on replication.failover) Environment variable: TT_REPLICATION_ELECTION_MODE

replication.election_timeout

Specifies the timeout (in seconds) between election rounds in the leader election process if the previous round ended up with a split vote.

It is quite big, and for most of the cases, it can be lowered to 300-400 ms.

To avoid the split vote repeat, the timeout is randomized on each node during every new election, from 100% to 110% of the original timeout value. For example, if the timeout is 300 ms and there are 3 nodes started the election simultaneously in the same term, they can set their election timeouts to 300, 310, and 320 respectively, or to 305, 302, and 324, and so on. In that way, the votes will never be split because the election on different nodes won't be restarted simultaneously.

Type: number

Default: 5

Environment variable: TT_REPLICATION_ELECTION_TIMEOUT

replication.election_fencing_mode

Specifies the leader fencing mode that affects the leader election process. When the parameter is set to soft or strict, the leader resigns its leadership if it has less than replication.synchro_quorum of alive connections to the cluster nodes. The resigning leader receives the status of a follower in the current election term and becomes read-only.

In soft mode, a connection is considered dead if there are no responses for 4 * replication.timeout seconds both on the current leader and the followers.
In strict mode, a connection is considered dead if there are no responses for 2 * replication.timeout seconds on the current leader and 4 * replication.timeout seconds on the followers. This improves the chances that there is only one leader at any time.

Fencing applies to the instances that have the replication.election_mode set to candidate or manual. To turn off leader fencing, set election_fencing_mode to off.

Type: string

Default: soft

Possible values: off`, `soft`, `strict

Environment variable: TT_REPLICATION_ELECTION_FENCING_MODE

replication.failover

A failover mode used to take over a master role when the current master instance fails. The following modes are available:

off

Leadership in a replica set is controlled using the database.mode option. In this case, you can set the database.mode option to rw on all instances in a replica set to make a master-master configuration.

The default database.mode is determined as follows: rw if there is one instance in a replica set; ro if there are several instances.
manual

Leadership in a replica set is controlled using the (replicaset_name.leader ) option. In this case, a master-master configuration is forbidden.

In the manual mode, the database.mode option cannot be set explicitly. The leader is configured in the read-write mode, all the other instances are read-only.
election

Automated leader election is used to control leadership in a replica set.

In the election mode, database.mode and (replicaset_name.leader ) shouldn't be set explicitly.
supervised (Enterprise Edition only)

Leadership in a replica set is controlled using an external failover coordinator.

In the supervised mode, database.mode and (replicaset_name.leader ) shouldn't be set explicitly.

replication.peers

URIs of instances that constitute a replica set. These URIs are used by an instance to connect to another instance as a replica.

Alternatively, you can use iproto.advertise.peer to specify a URI used to advertise the current instance to other cluster members.

Example

In the example below, the following configuration options are specified:

In the credentials section, the replicator user with the replication role is created.
replication.peers specifies URIs of replica set instances.


credentials:  users:    replicator:      password: 'topsecret'      roles: [replication]replication:  peers:  - replicator:topsecret@127.0.0.1:3301  - replicator:topsecret@127.0.0.1:3302  - replicator:topsecret@127.0.0.1:3303

Type: array

Default: [box.NULL](../reference_lua/box/box_null#box-null)

Environment variable: TT_REPLICATION_PEERS

replication.skip_conflict

By default, if a replica adds a unique key that another replica has added, replication stops with the ER_TUPLE_FOUND error. If replication.skip_conflict is set to true, such errors are ignored.

Type: boolean

Default: false

Environment variable: TT_REPLICATION_SKIP_CONFLICT

replication.sync_lag

The maximum delay (in seconds) between the time when data is written to the master and the time when it is written to a replica. If replication.sync_lag is set to nil or 365 * 100 * 86400 (TIMEOUT_INFINITY), a replica is always considered to be "synced".

Type: number

Default: 10

Environment variable: TT_REPLICATION_SYNC_LAG

replication.sync_timeout

The timeout (in seconds) that a node waits when trying to sync with other nodes in a replica set after connecting or during a configuration update. This could fail indefinitely if replication.sync_lag is smaller than network latency, or if the replica cannot keep pace with master updates. If replication.sync_timeout expires, the replica enters orphan status.

Type: number

Default: 0

Environment variable: TT_REPLICATION_SYNC_TIMEOUT

replication.synchro_queue_max_size

Since: 3.3.0

The maximum size of the synchronous transaction queue on a master node, in bytes. The size limit isn't strict, i.e. if there's at least one free byte, the whole write request fits and no blocking is involved. This parameter ensures that the queue does not grow indefinitely, potentially impacting performance and resource usage, and applies only to the master node.

The 0 value disables the limit.

If the synchronous queue reaches the configured size limit, new transactions attempting to enter the queue are discarded. In such cases, the system returns an error to the user: The synchronous transaction queue is full.

This size limitation does not apply during the recovery process. Transactions processed during recovery are unaffected by the queue size limit.

Use the following command to view the current size of the synchronous queue:


box.info.synchro.queue.size


tarantool> box.info.synchro---- queue:    owner: 1    size: 60    busy: false    len: 1    term: 2    quorum: 2...

Set the synchronous queue size limit in the configuration file:


replication:    synchro_queue_max_size: 33554432  # Limit set to 32 MB

Type: integer

Default: 16777216 (16 MB)

Environment variable: TT_REPLICATION_SYNCHRO_QUEUE_MAX_SIZE

replication.synchro_quorum

A number of replicas that should confirm the receipt of a synchronous transaction before it can finish its commit.

This option supports dynamic evaluation of the quorum number. For example, the default value is N / 2 + 1 where N is the current number of replicas registered in a cluster. Once any replicas are added or removed, the expression is re-evaluated automatically.

Note that the default value (at least 50% of the cluster size + 1) guarantees data reliability. Using a value less than the canonical one might lead to unexpected results, including a split-brain.

replication.synchro_quorum is not used on replicas. If the master fails, the pending synchronous transactions will be kept waiting on the replicas until a new master is elected.

Type: string, number

Default: N / 2 + 1

Environment variable: TT_REPLICATION_SYNCHRO_QUORUM

replication.synchro_timeout

For synchronous replication only. Specify how many seconds to wait for a synchronous transaction quorum replication until it is declared failed and is rolled back.

It is not used on replicas, so if the master fails, the pending synchronous transactions will be kept waiting on the replicas until a new master is elected.

Type: number

Default: 5

Environment variable: TT_REPLICATION_SYNCHRO_TIMEOUT

replication.threads

The number of threads spawned to decode the incoming replication data.

In most cases, one thread is enough for all incoming data. Possible values range from 1 to 1000. If there are multiple replication threads, connections to serve are distributed evenly between the threads.

Type: integer

Default: 1

Environment variable: TT_REPLICATION_THREADS

replication.timeout

A time interval (in seconds) used by a master to send heartbeat requests to a replica when there are no updates to send to this replica. For each request, a replica should return a heartbeat acknowledgment.

If a master or replica gets no heartbeat message for 4 * replication.timeout seconds, a connection is dropped and a replica tries to reconnect to the master.

roles

This section describes configuration parameters related to application roles.

roles

Specify the roles of an instance. To specify a role's configuration, use the roles_cfg option.

Type: array

Default: nil

Environment variable: TT_ROLES

roles_cfg

Specify a role's configuration. This option accepts a role name as the key and a role's configuration as the value. To specify the roles of an instance, use the roles option.

Type: map

Default: nil

Environment variable: TT_ROLES_CFG

security

The security section defines configuration parameters related to various security settings.

security.auth_delay

Specify a period of time (in seconds) that a specific user should wait for the next attempt after failed authentication.

The security.auth_retries option lets a client try to authenticate the specified number of times before security.auth_delay is enforced.

In the configuration below, Tarantool lets a client try to authenticate with the same username three times. At the fourth attempt, the authentication delay configured with security.auth_delay is enforced. This means that a client should wait 10 seconds after the first failed attempt.


security:  auth_delay: 10  auth_retries: 2

Type: number

Default: 0

Environment variable: TT_SECURITY_AUTH_DELAY

security.auth_retries

Specify the maximum number of authentication retries allowed before security.auth_delay is enforced. The default value is 0, which means security.auth_delay is enforced after the first failed authentication attempt.

The retry counter is reset after security.auth_delay seconds since the first failed attempt. For example, if a client tries to authenticate fewer than security.auth_retries times within security.auth_delay seconds, no authentication delay is enforced. The retry counter is also reset after any successful authentication attempt.

Type: integer

Default: 0

Environment variable: TT_SECURITY_AUTH_RETRIES

security.auth_type

Specify a protocol used to authenticate users. The possible values are:

chap-sha1: use the CHAP protocol with SHA-1 hashing applied to passwords.
pap-sha256: use PAP authentication with the SHA256 hashing algorithm.

Note that CHAP stores password hashes in the _user space unsalted. If an attacker gains access to the database, they may crack a password, for example, using a rainbow table. For PAP, a password is salted with a user-unique salt before saving it in the database, which keeps the database protected from cracking using a rainbow table.

To enable PAP, specify the security.auth_type option as follows:


security:  auth_type: 'pap-sha256'

sharding.bucket_count

The total number of buckets in a cluster. Learn more in vshard_config_bucket_count.

Example


sharding:  bucket_count: 1000

Type: integer

Default: 3000

Environment variable: TT_SHARDING_BUCKET_COUNT

sharding.discovery_mode

A mode of the background discovery fiber used by the router to find buckets. Learn more in vshard.router.discovery_set().

Type: string

Default: 'on'

Possible values: 'on', 'off', 'once'

Environment variable: TT_SHARDING_DISCOVERY_MODE

sharding.failover_ping_timeout

The timeout (in seconds) after which a node is considered unavailable if there are no responses during this period. The failover fiber is used to detect if a node is down.

Type: number

Default: 5

Environment variable: TT_SHARDING_FAILOVER_PING_TIMEOUT

sharding.lock

Whether a replica set is locked. A locked replica set cannot receive new buckets nor migrate its own buckets.

Type: boolean

Default: nil

Environment variable: TT_SHARDING_LOCK

sharding.rebalancer_disbalance_threshold

The maximum bucket disbalance threshold (in percent). The disbalance is calculated for each replica set using the following formula:


|etalon_bucket_count - real_bucket_count| / etalon_bucket_count * 100

Type: number

Default: 1

Environment variable: TT_SHARDING_REBALANCER_DISBALANCE_THRESHOLD

sharding.rebalancer_max_receiving

The maximum number of buckets that can be received in parallel by a single replica set. This number must be limited because the rebalancer sends a large number of buckets from the existing replica sets to the newly added one. This produces a heavy load on the new replica set.

Example

Suppose, rebalancer_max_receiving is equal to 100 and bucket_count is equal to 1000. There are 3 replica sets with 333, 333, and 334 buckets on each respectively. When a new replica set is added, each replica set's etalon_bucket_count becomes equal to 250. Rather than receiving all 250 buckets at once, the new replica set receives 100, 100, and 50 buckets sequentially.

Type: integer

Default: 100

Environment variable: TT_SHARDING_REBALANCER_MAX_RECEIVING

sharding.rebalancer_max_sending

The degree of parallelism for parallel rebalancing.

Type: integer

Default: 1

Maximum: 15

Environment variable: TT_SHARDING_REBALANCER_MAX_SENDING

sharding.rebalancer_mode

Since: 3.1.0

Configure how a rebalancer is selected:

auto (default): if there are no replica sets with the rebalancer sharding role (sharding.roles), a replica set with the rebalancer is selected automatically among all replica sets.
manual: one of the replica sets should have the rebalancer sharding role. The rebalancer is in this replica set.
off: rebalancing is turned off regardless of whether a replica set with the rebalancer sharding role exists or not.

Type: string

Default: 'auto'

Environment variable: TT_SHARDING_REBALANCER_MODE

sharding.roles

Roles of a replica set in regard to sharding. A replica set can have the following roles:

router: a replica set acts as a router.
storage: a replica set acts as a storage.
rebalancer: a replica set acts as a rebalancer.

The rebalancer role is optional. If it is not specified, a rebalancer is selected automatically from the master instances of replica sets.

There can be at most one replica set with the rebalancer role. Additionally, this replica set should have a storage role.

Example


replicasets:  storage-a:    sharding:      roles: [storage, rebalancer]

sharding.sched_move_quota

A scheduler's bucket move quota used by the rebalancer.

sched_move_quota defines how many bucket moves can be done in a row if there are pending storage refs. Then, bucket moves are blocked and a router continues making map-reduce requests.

sharding.sched_ref_quota

A scheduler's storage ref quota used by a router's map-reduce API. For example, the vshard.router.map_callrw() function implements consistent map-reduce over the entire cluster.

sched_ref_quota defines how many storage refs, therefore map-reduce requests, can be executed on the storage in a row if there are pending bucket moves. Then, storage refs are blocked and the rebalancer continues bucket moves.

sharding.shard_index

The name or ID of a TREE index over the bucket id. Spaces without this index do not participate in a sharded Tarantool cluster and can be used as regular spaces if needed. It is necessary to specify the first part of the index, other parts are optional.

sharding.sync_timeout

The timeout to wait for synchronization of the old master with replicas before demotion. Used when switching a master or when manually calling the sync() function.

Type: number

Default: 1

Environment variable: TT_SHARDING_SYNC_TIMEOUT

sharding.weight

Since: 3.1.0

The relative amount of data that a replica set can store. Learn more at vshard-replica-set-weights.

Type: number

Default: 1

Environment variable: TT_SHARDING_WEIGHT

sharding.zone

A zone that can be set for routers and replicas. This allows sending read-only requests not only to a master instance but to any available replica that is the nearest to the router.

Type: integer

Default: nil

Environment variable: TT_SHARDING_ZONE

snapshot

The snapshot section defines configuration parameters related to the snapshot files. To learn more about the snapshots' configuration, check the Persistence page.

snapshot.dir

A directory where memtx stores snapshot (.snap) files. A relative path in this option is interpreted as relative to process.work_dir.

By default, snapshots and WAL files are stored in the same directory. However, you can set different values for the snapshot.dir and wal.dir options to store them on different physical disks for performance matters.

Type: string

Default: 'var/lib/{{ instance_name }}'

Environment variable: TT_SNAPSHOT_DIR

snapshot.snap_io_rate_limit

Reduce the throttling effect of box.snapshot() on INSERT/UPDATE/DELETE performance by setting a limit on how many megabytes per second it can write to disk. The same can be achieved by splitting wal.dir and snapshot.dir locations and moving snapshots to a separate disk. The limit also affects what box.stat.vinyl().regulator may show for the write rate of dumps to .run and .index files.

Type: number

Default: box.NULL

Environment variable: TT_SNAPSHOT_SNAP_IO_RATE_LIMIT

snapshot.count

The maximum number of snapshots that are stored in the snapshot.dir directory. If the number of snapshots after creating a new one exceeds this value, the Tarantool garbage collector deletes old snapshots. If snapshot.count is set to zero, the garbage collector does not delete old snapshots.

Example

In the example, the checkpoint daemon creates a snapshot every two hours until it has created three snapshots. After creating a new snapshot (the fourth one), the oldest snapshot and any associated write-ahead-log files are deleted.


snapshot:  by:    interval: 7200  count: 3

Type: integer

Default: 2

Environment variable: TT_SNAPSHOT_COUNT

snapshot.by.*

snapshot.by.interval

The interval in seconds between actions by the checkpoint daemon. If the option is set to a value greater than zero, and there is activity that causes change to a database, then the checkpoint daemon calls box.snapshot() every snapshot.by.interval seconds, creating a new snapshot file each time. If the option is set to zero, the checkpoint daemon is disabled.

Example

In the example, the checkpoint daemon creates a new database snapshot every two hours, if there is activity.


by:  interval: 7200

Type: number

Default: 3600

Environment variable: TT_SNAPSHOT_BY_INTERVAL

snapshot.by.wal_size

The threshold for the total size in bytes for all WAL files created since the last snapshot taken. Once the configured threshold is exceeded, the WAL thread notifies the checkpoint daemon that it must make a new snapshot and delete old WAL files.

Type: integer

Default: 10\^18

Environment variable: TT_SNAPSHOT_BY_WAL_SIZE

sql

The sql section defines configuration parameters related to SQL.

sql.cache_size

sql.cache_size

The maximum cache size (in bytes) for all SQL prepared statements. To see the actual cache size, use box.info.sql().cache.size.

Type: integer

Default: 5242880

Environment variable: TT_SQL_CACHE_SIZE

vinyl

The vinyl section defines configuration parameters related to the vinyl storage engine.

vinyl.bloom_fpr

A bloom filter's false positive rate – the suitable probability of the bloom filter to give a wrong result. The vinyl.bloom_fpr setting is a default value for the bloom_fpr option passed to space_object:create_index().

Type: number

Default: 0.05

Environment variable: TT_VINYL_BLOOM_FPR

vinyl.cache

The cache size for the vinyl storage engine. The cache can be resized dynamically.

Type: integer

Default: 128 * 1024 * 1024

Environment variable: TT_VINYL_CACHE

vinyl.defer_deletes

Enable the deferred DELETE optimization in vinyl. It was disabled by default since Tarantool version 2.10 to avoid possible performance degradation of secondary index reads.

Type: boolean

Default: false

Environment variable: TT_VINYL_DEFER_DELETES

vinyl.dir

A directory where vinyl files or subdirectories will be stored.

This option may contain a relative file path. In this case, it is interpreted as relative to process.work_dir.

Type: string

Default: 'var/lib/{{ instance_name }}'

Environment variable: TT_VINYL_DIR

vinyl.max_tuple_size

The size of the largest allocation unit, for the vinyl storage engine. It can be increased if it is necessary to store large tuples.

Type: integer

Default: 1024 * 1024

Environment variable: TT_VINYL_MAX_TUPLE_SIZE

vinyl.memory

The maximum number of in-memory bytes that vinyl uses.

Type: integer

Default: 128 * 1024 * 1024

Environment variable: TT_VINYL_MEMORY

vinyl.page_size

The page size. A page is a read/write unit for vinyl disk operations. The vinyl.page_size setting is a default value for the page_size option passed to space_object:create_index().

Type: integer

Default: 8 * 1024

Environment variable: TT_VINYL_PAGE_SIZE

vinyl.range_size

The default maximum range size for a vinyl index, in bytes. The maximum range size affects the decision of whether to split a range.

The wal section defines configuration parameters related to write-ahead log. To learn more about the WAL configuration, check the Persistence page.

wal.cleanup_delay

The delay in seconds used to prevent the Tarantool garbage collector from immediately removing write-ahead log files after a node restart. This delay eliminates possible erroneous situations when the master deletes WALs needed by replicas after restart. As a consequence, replicas sync with the master faster after its restart and don't need to download all the data again. Once all the nodes in the replica set are up and running, a scheduled garbage collection is started again even if wal.cleanup_delay has not expired.

wal.dir

A directory where write-ahead log (.xlog) files are stored. A relative path in this option is interpreted as relative to process.work_dir.

By default, WAL files and snapshots are stored in the same directory. However, you can set different values for the wal.dir and snapshot.dir options to store them on different physical disks for performance matters.

Type: string

Default: 'var/lib/{{ instance_name }}'

Environment variable: TT_WAL_DIR

wal.dir_rescan_delay

The time interval in seconds between periodic scans of the write-ahead-log file directory, when checking for changes to write-ahead-log files for the sake of replication or hot standby.

Type: number

Default: 2

Environment variable: TT_WAL_DIR_RESCAN_DELAY

wal.max_size

The maximum number of bytes in a single write-ahead log file. When a request would cause an .xlog file to become larger than wal.max_size, Tarantool creates a new WAL file.

Type: integer

Default: 268435456

Environment variable: TT_WAL_MAX_SIZE

wal.mode

Specify fiber-WAL-disk synchronization mode as:

none: write-ahead log is not maintained. A node with wal.mode set to none can't be a replication master.
write: fibers wait for their data to be written to the write-ahead log (no fsync(2)).
fsync: fibers wait for their data, fsync(2) follows each write(2).

Type: string

Default: 'write'

Environment variable: TT_WAL_MODE

wal.queue_max_size

The size of the queue in bytes used by a replica to submit new transactions to a write-ahead log (WAL). This option helps limit the rate at which a replica submits transactions to the WAL.

Limiting the queue size might be useful when a replica is trying to sync with a master and reads new transactions faster than writing them to the WAL.

Type: integer

Default: 16777216

Environment variable: TT_WAL_QUEUE_MAX_SIZE

wal.retention_period

Since: 3.1.0 (Enterprise Edition only)

The delay in seconds used to prevent the Tarantool garbage collector from removing a write-ahead log file after it has been closed. If a node is restarted, wal.retention_period counts down from the last modification time of the write-ahead log file.

The garbage collector doesn't track write-ahead logs that are to be relayed to anonymous replicas, such as:

Anonymous replicas added as a part of a cluster configuration (see replication.anon).
CDC (Change Data Capture) that retrieves data using anonymous replication.

In case of a replica or CDC downtime, the required write-ahead logs can be removed. As a result, such a replica needs to be rebootstrapped. You can use wal.retention_period to prevent such issues.

Note that wal.cleanup_delay option also sets the delay used to prevent the Tarantool garbage collector from removing write-ahead logs. The difference is that the garbage collector doesn't take into account wal.cleanup_delay if all the nodes in the replica set are up and running, which may lead to the removal of the required write-ahead logs.

Type: number

Default: 0

Environment variable: TT_WAL_RETENTION_PERIOD

wal.ext.*

This section describes options related to WAL extensions.

wal.ext.new

Enable storing a new tuple for each CRUD operation performed. The option is in effect for all spaces. To adjust the option for specific spaces, use the wal.ext.spaces option.

Type: boolean

Default: false

Environment variable: TT_WAL_EXT_NEW

wal.ext.old

Enable storing an old tuple for each CRUD operation performed. The option is in effect for all spaces. To adjust the option for specific spaces, use the wal.ext.spaces option.

Type: boolean

Default: false

Environment variable: TT_WAL_EXT_OLD

wal.ext.spaces

Enable or disable storing an old and new tuple in the WAL record for a given space explicitly. The configuration for specific spaces has priority over the configuration in the wal.ext.new and wal.ext.old options.

The option is a key-value pair:

The key is a space name (string).
The value is a table that includes two optional boolean options: old and new. The format and the default value of these options are described in wal.ext.old and wal.ext.new.

Example

In the example, only new tuples are added to the log for the bands space.


ext:  new: true  old: true  spaces:    bands:      old: false

Type: map

Default: nil

Environment variable: TT_WAL_EXT_SPACES