Got alerted by Contegix.

      We received alerts for confluence.atlassian.com at 05:55 CST on Friday, February 22nd. At that time, the instance was displaying a resin error, and I noted several dozen "idle in transaction" postgres connections. I have captured a ps of the pg connections, taken thread dumps, and attached the logs. I have restarted the instance following, and the site appears good at this time.

      I looked at the access logs, but could not see any obvious patterns. We decided to track this systematically by regularly quiering the database for what statements the idle in transaction connections were trying to execute, to find the root cause.

      See http://wiki.dspace.org/index.php/Idle_In_Transaction_Problem for an explanation

      We need to investigate further.

          Form Name

            [CONFSERVER-10798] Database connections end up in "idle-in-transaction"-mode

            No work has yet been logged on this issue.

              matt@atlassian.com Matt Ryall
              pfragemann Per Fragemann [Atlassian]
              Affected customers:
              0 This affects my team
              Watchers:
              0 Start watching this issue

                Created:
                Updated:
                Resolved: