Samuel Smith schrieb am 04.06.2020 um 21:59:
Sorry, I should have clarified that I was aware of the pg_stat_activity table. That is how we found the problem in the first place. And yes I could just write a bash script and run it in cron. I just didn't know if there was a more "official" way to go about this since it is probably a common monitoring point and/or if something like this was already made.
What about setting idle_in_transaction_session_timeout to some reasonably high value (15 minutes?) Then you would get an error in your application because the connection was terminated and you would learn early about the problem. Even with monitoring enabled, you would probably still kill those sessions manually as the application most probably can't commit them properly any more. Thomas
