The fix is to restart postgres ... If I only close the connections the problem is still these so I need to restart postgres.
How are you closing the connections?
I restart the application server. The problem is that the max_idle connections was set to 1000 on jdbc connection so once the spike happened the app would run with 300 connections and 250 of them or so IDLE for most of the time. I am fixing that
Hi Ionana, thanks for the responses. Does restarting the app server successfully cause all of those connections to terminate? If so, and yet you still have memory problems, then there is still the mystery of where the memory is going.