Patroni cluster management commands v23

Patroni can be used as a single master failover manager with the M1 architecture using the following command options.

tpaexec configure cluster_name -a M1 --enable-patroni --postgresql 14

Or by setting the config.yml option

cluster_vars:
  failover_manager: patroni

TPA configure will add 3 etcd nodes and 2 haproxy nodes. Etcd is used for the Distributed Configuration Store (DCS). Patroni supports other DCS backends, but they are not currently supported by EDB or TPA.

TPA uses Patroni's feature of converting an existing PostgreSQL standalone server. This allows for TPA to initialise and manage configuration. Once a single PostgreSQL server and database has been created, Patroni will create replicas and configure replication. TPA will then remove any postgres configuration files used during setup.

Once set up, Postgres can continue to be managed using TPA and settings in config.yml for the cluster. You can also use Patroni interfaces, such as the command line patronictl and the REST API, but it is recommended to use TPA methods wherever possible.

These configuration variables can be used to control certain behaviours in the deployment of Patroni in TPA.

Patroni configuration file settings

Configuration for patroni is built from three layers, starting with defaults set by the Patroni daemon, config loaded from the DCS, and finally from local configuration. The last can be controlled from either configuration file and overrides via the environment. TPA controls the configuration file and values are built up in this order.

DCS config to be sent to the API and stored in the bootstrap section of the config file:

• TPA vars for postgres are loaded into the DCS settings, see postgresql.conf.md. Some features are not supported, see notes below.
• Patroni defaults for DCS settings
• User supplied defaults in patroni_dynamic_conf_settings, if you want to override any DCS settings you can do that here.

Local config stored in the YAML configuration file:

• bootstrap.dcs loaded from previous steps above.
• configuration enabled by feature flags, such as patroni_ssl_enabled, see table above.
• then finally overloaded from user supplied settings, the patroni_conf_settings option. If you want to change or add configuration not controlled by a feature flag then this is the best place to do it.

Please note that configuration is merged on top of configuration generated by TPA from cluster information, such as IP addresses, port numbers, cluster roles, etc. Exercise caution in what you override as this might affect the stable operation of the cluster.

As Patroni stores all postgres configuration in the DCS and controls how and when this is distributed to postgres, some features of TPA are incompatible with patroni:

• It is not possible to change the template used to generate postgresql.conf with the setting postgres_conf_template.
• You cannot change the location of Postgres config files with the setting postgres_conf_dir.

Patroni configuration in TPA config.yml

You can override single values:

cluster_vars:
  patroni_conf_settings:
    bootstrap:
      dcs:
        ttl: 120

Or full blocks (with an example from Patroni docs):

cluster_vars:
  patroni_conf_settings:
    restapi:
      http_extra_headers:
        'X-Frame-Options': 'SAMEORIGIN'
        'X-XSS-Protection': '1; mode=block'
        'X-Content-Type-Options': 'nosniff'
      https_extra_headers:
        'Strict-Transport-Security': 'max-age=31536000; includeSubDomains'

If you want to negate a value or section that is present in the default TPA config vars you can set the value to null. This will cause patroni to ignore this section when loading the config file.

For example the default TPA config for log is

log:
  dir: /var/log/patroni

To turn off logging add this to config.yml:

cluster_vars:
  patroni_conf_settings:
    log: null

TPA provides these minimal set of tools for managing Patroni clusters.

Status

To see the current status of the TPA cluster according to Patroni run

tpaexec status cluster_name

Switchover

To perform a switchover to a replica node (e.g. to perform maintenance) run the command

tpaexec switchover cluster_name new_primary

The new_primary argument must be the name of an existing cluster node that is currently running as a healthy replica. Checks will be performed to ensure this is true before a switchover is performed.

Once a switchover has been performed it is recommended that you run deploy and test to ensure a healthy cluster.

tpaexec deploy cluster_name
tpaexec test cluster_name

TPA will detect the current role of nodes during deploy regardless of what config.yml contains, for example if a different node is the leader.

• On this page
• Patroni configuration file settings
• Status
• Switchover