PgBouncer server conn crashed Error with Jira DC on PostgreSQL HA

[Dave King]

Hi Atlassian Community,

We’re running Jira Data Center with a PostgreSQL high-availability setup (Patroni + PgBouncer.HaProxy), and we’re encountering the following intermittent error in our logs:

May 05 04:12:09 jira-node-01 pgbouncer[56009]: C-0x55e185483240: jiraprod/jirauser@127.0.0.1:54248 pooler error: server conn crashed?

This seems to happen sporadically, and we can’t yet determine what’s causing the server connection to break.

What can typically cause PgBouncer to log server conn crashed? — is this a symptom of Postgres backend crash/restart, query issues, or something Jira could be doing?

Any insights  would be greatly appreciated.

Thanks in advance!

[Mercy]

That PgBouncer message indicates that an existing server connection in its pool was terminated unexpectedly while Jira was still using it. In a PostgreSQL HA environment managed by Patroni, this usually happens when the primary node is being restarted or failover events briefly invalidate active connections, causing PgBouncer to drop them as “crashed.” Jira itself doesn’t explicitly kill backend connections, but long-running queries or idle transactions can be affected by Patroni’s health checks or timeouts. When PgBouncer retries, Jira logs can show transient database errors such as “connection reset by peer,” which align with these backend interruptions.