We have a few servers where we see close_wait connections piling up until we reached the limit. As I understood that the application should close the close_wait sockets I suspect this to be an application issues. I can't find anything in the code execution indicating that nor do I something suspicious in Dynatrace that would warn us.
Solved! Go to Solution.
you could put in a custom detection for that string of close_warn and Dynatrace will alert you. This method can also be done at a log standpoint as well.
Close_wait()'s can pile up when a TCP connection gets closed from the other end and leaves your end dangling. So perhaps it would be helpful to track the call tree of the close wait threads AND track the remote endpoint. This might give you visibility into a pattern of loosing connections to a common remote endpoint which is really the root cause.