Managing cluster configurations

Publishing local configurations

tt cluster publish can modify local cluster configurations stored in config.yaml files inside application directories.

To write a configuration to a local config.yaml, run tt cluster publish with two arguments:

• the application name.
• the path to a YAML file from which the configuration should be taken.

$ tt cluster publish myapp source.yaml 

Publishing configurations in centralized storages

tt cluster publish can modify centralized cluster configurations in storages of both supported types: etcd or a Tarantool-based configuration storage.

To publish a configuration from a file to a centralized configuration storage, run tt cluster publish with a URI of this storage’s instance as the target. For example, the command below publishes a configuration from source.yaml to a local etcd instance running on the default port 2379:

$ tt cluster publish "http://localhost:2379/myapp" source.yaml 

A URI must include a prefix that is unique for the application. It can also include credentials and other connection parameters. Find the detailed description of the URI format in URI format.

Publishing configurations of specific instances

In addition to whole cluster configurations, tt cluster publish can manage configurations of specific instances within applications: rewrite configurations of existing instances and add new instance configurations.

In this case, it operates with YAML fragments that describe a single instance configuration section. For example, the following YAML file can be a source when publishing an instance configuration:

# instance_source.yaml iproto:  listen:  - uri: 127.0.0.1:3311 

To send an instance configuration to a local config.yaml, run tt cluster publish with the application:instance pair as the target argument:

$ tt cluster publish myapp:instance-002 instance_source.yaml 

To send an instance configuration to a centralized configuration storage, specify the instance name in the name argument of the storage URI:

$ tt cluster publish "http://localhost:2379/myapp?name=instance-002" instance_source.yaml 

If the instance already exists, this call overwrites its configuration with the one from the file.

To add a new instance configuration from a YAML fragment, specify the name to assign to the new instance and its location in the cluster topology – replica set and group – in the --replicaset and --group options.

To add a new instance instance-003 to the replicaset-001 replica set:

$ tt cluster publish "http://localhost:2379/myapp?name=instance-003" instance_source.yaml --replicaset replicaset-001 

Configuration validation

tt cluster publish validates configurations against the Tarantool configuration schema and aborts in case of an error. To skip the validation, add the --force option:

$ tt cluster publish myapp source.yaml --force 

Publishing configurations with integrity check

When called with the --with-integrity-check option, tt cluster publish generates a checksum of the configurations it publishes. It signs the checksum using the private key passed as the option argument, and writes it into the configuration store.

$ tt cluster publish "http://localhost:2379/myapp" source.yaml --with-integrity-check private.pem 

If an application configuration is published this way, it can be checked for integrity using the --integrity-check global option.

$ tt --integrity-check public.pem cluster show myapp $ tt --integrity-check public.pem start myapp 

Learn more about integrity checks upon application startup and in runtime in the tt start reference.

To ensure the configuration integrity when updating it, call tt cluster publish with two options:

• --integrity-check PUBLIC_KEY global option checks that the configuration wasn’t changed since it was published
• --with-integrity-check PRIVATE_KEY generates new hash and signature for future integrity checks of the updated configuration.

$ tt --integrity-check public.pem cluster publish \  --with-integrity-check private.pem \  "http://localhost:2379/myapp" source.yaml 

Displaying local configurations

tt cluster show can read local cluster configurations stored in config.yaml files inside application directories.

To print a local configuration from an application’s config.yaml, specify the application name as an argument:

$ tt cluster show myapp 

Displaying configurations from centralized storages

tt cluster show can display centralized cluster configurations from configuration storages of both supported types: etcd or a Tarantool-based configuration storage.

To print a cluster configuration from a centralized storage, run tt cluster show with a storage URI including the prefix identifying the application. For example, to print myapp’s configuration from a local etcd storage:

$ tt cluster show "http://localhost:2379/myapp" 

Displaying configurations of specific instances

In addition to whole cluster configurations, tt cluster show can display configurations of specific instances within applications. In this case, it prints YAML fragments that describe a single instance configuration section.

To print an instance configuration from a local config.yaml, use the application:instance argument:

$ tt cluster show myapp:instance-002 

To print an instance configuration from a centralized configuration storage, specify the instance name in the name argument of the URI:

$ tt cluster show "http://localhost:2379/myapp?name=instance-002" 

Configuration validation

To validate configurations when printing them with tt cluster show, enable the validation by adding the --validate option:

$ tt cluster show "http://localhost:2379/myapp" --validate 

promote

$ tt cluster replicaset promote CONFIG_URI INSTANCE_NAME [OPTION ...] 

tt cluster replicaset promote promotes the specified instance, making it a leader of its replica set. This command works on Tarantool clusters with centralized configuration and with failover modes off and manual. It updates the centralized configuration according to the specified arguments and reloads it:

• off failover mode: the command sets database.mode to rw on the specified instance.

  Important

  If failover is off, the command doesn’t consider the modes of other replica set members, so there can be any number of read-write instances in one replica set.

• manual failover mode: the command updates the leader option of the replica set configuration. Other instances of this replica set become read-only.

Example:

$ tt cluster replicaset promote "http://localhost:2379/myapp" storage-001-a 

demote

$ tt cluster replicaset demote CONFIG_URI INSTANCE_NAME [OPTION ...] 

tt cluster replicaset demote demotes an instance in a replica set. This command works on Tarantool clusters with centralized configuration and with failover mode off.

The command sets the instance’s database.mode to ro and reloads the configuration.

expel

$ tt cluster replicaset expel CONFIG_URI INSTANCE_NAME [OPTION ...] 

tt cluster replicaset expel expels an instance from the cluster. Example:

$ tt cluster replicaset expel "http://localhost:2379" storage-b-002 

roles

$ tt cluster replicaset roles [add|remove] CONFIG_URI ROLE_NAME [OPTION ...] 

tt cluster replicaset roles manages application roles in the configuration scope specified in the command options. It has two subcommands:

• add adds a role
• remove removes a role

Use the --global, --group, --replicaset, --instance options to specify the configuration scope to add or remove roles. For example, to add a role to all instances in a replica set:

$ tt cluster replicaset roles add "http://localhost:2379" roles.my-role --replicaset storage-a 

To remove a role defined in the global configuration scope:

$ tt cluster replicaset roles remove "http://localhost:2379" roles.my-role --global 

Implementation details

The changes that tt cluster replicaset makes to the configuration storage occur transactionally. Each call creates a new revision. In case of a revision mismatch, an error is raised.

If the cluster configuration is distributed over multiple keys in the configuration storage (for example, in two paths /myapp/config/k1 and /myapp/config/k2), the affected instance configuration can be present in more that one of them. If it is found under several different keys, the command prompts the user to choose a key for patching. You can skip the selection by adding the -f/--force option:

$ tt cluster replicaset promote "http://localhost:2379/myapp" storage-001-a --force 

In this case, the command selects the key for patching automatically. A key’s priority is determined by the detail level of the instance or replica set configuration stored under this key. For example, when failover is off, a key with instance.database options takes precedence over a key with the only instance field. In case of equal priority, the first key in the lexicographical order is patched.

switch

$ tt cluster failover switch CONFIG_URI INSTANCE_NAME [OPTION ...] 

tt cluster failover switch appoints the specified instance to be a master. This command accepts the following arguments and options:

• CONFIG_URI: A URI of the cluster configuration storage.
• INSTANCE_NAME: An instance name.
• [OPTION ...]: Options to pass to the command.

In the example below, tt cluster failover switch appoints storage-a-002 to be a master:

$ tt cluster failover switch http://localhost:2379/myapp storage-a-002 To check the switching status, run: tt cluster failover switch-status http://localhost:2379/myapp b1e938dd-2867-46ab-acc4-3232c2ef7ffe 

Note that the command output includes an identifier of the task responsible for switching a master. You can use this identifier to see the status of switching a master instance using tt cluster failover switch-status.

switch-status

$ tt cluster failover switch-status CONFIG_URI TASK_ID 

tt cluster failover switch-status shows the status of switching a master instance. This command accepts the following arguments:

• CONFIG_URI: A URI of the cluster configuration storage.
• TASK_ID: An identifier of the task used to switch a master instance. You can find the task identifier in the tt cluster failover switch command output.

Example:

$ tt cluster failover switch-status http://localhost:2379/myapp b1e938dd-2867-46ab-acc4-3232c2ef7ffe