Worker failover support for citus community version using pg_auto_failover.
This is a simple service to monitor changes in a pg_auto_failover monitor and update workers' addresses in citus coordinator. The community version of citus does not support failover so you have to update the workers manually in case of a failover.
It will check both the citus coordinator and pg_auto_failover monitor and if it finds a worker node that is not primary, it will replace it with the primary one.
You need to provide the connection details for both your coordinator and your monitor. Note that the coordinator should also be monitored by pg_auto_failover and you should provide the formation that is responsible for your coordinator, the primary node then will be found and used to update the workers.
You need to provide a config file. Check config.example.yml for a sample config.
Monitor: PSQL connection details for your monitor.
Coordinator:
You need to provide the formation for your coordinator nodes as Formation instead of any host. It will be used to find the primary coordinator.
./citus-failover -f path/to/config.yml
Create a config file somewhere in your host. Then mount the path as /config to the container.
Map the port for API if enabled.
docker run -d -v /config/folder/path:/config -p 3002:3002 navid2zp/citus-failover:latest
A simple REST API is available. You can use it to check the state of your nodes. Set API.Enabled to true in your config file to enable the API. You also need to provide a secret string for the API. Then you'll need to provide this secret string in the header of your requests as SECRET to be authenticated and authorized.
Available endpoints:
/v1/nodes: List of all the nodes in your monitor.
/v1/coordinator: Primary coordinator address.
/v1/coordinators: List of all available coordinator nodes.
v/1/workers: List of primary worker nodes.