Description
The Zammad scheduler does not automatically recover its database connection following a PostgreSQL failover event. While the database becomes available again shortly after the failover, the scheduler remains in a failed state, continuously reporting ActiveRecord::ConnectionNotEstablished. A manual restart of the container/service resolves the issue immediately.
Error logs
E, [2026-02-18T01:27:36.234136#1-196240] ERROR -- : execute Channel.fetch (try_count 0) exited with error #<ActiveRecord::ConnectionNotEstablished:"connection to server at \"10.1.29.223\", port 5432 failed: server closed the connection unexpectedly\n\tThis probably means the server terminated abnormally\n\tbefore or while processing the request.\n"> in: 0.559769647 seconds.
E, [2026-02-18T01:27:36.357421#1-196240] ERROR -- : Can't reconnect to database #<ActiveRecord::ConnectionNotEstablished:"connection to server at \"10.1.29.223\", port 5432 failed: server closed the connection unexpectedly\n\tThis probably means the server terminated abnormally\n\tbefore or while processing the request.\n">
Environment
- Zammad 6.5.2 (image from 17.02.2026)
- Installation method: docker
- Database: Postgres 18 pg_autofailover through HAProxy
Our questions:
- Is there any settings that allows to configure connection pool and force recycling of DB connections ?
- Is there any effective container monitoring that can automate the restart of the zammad services ?