Open romanek-adam opened 4 years ago
Okay, I'm a bit surprised to see no additional comments on this report? We may be seeing problems with queues this in our production environment due to this!
Yes, the default could be changed for a quick fix. But the trigger should probably be updated with a configurable connection timeout.
Hi, this is a follow-up of my investigation related to an incident me and my team was hit yesterday after our Gerrit instance outage.
I described the whole story in Jenkins' bug tracker as it was Jenkins' Gerrit Trigger plugin which was affected by the issue. See JENKINS-33959 and my comment specifically.
I believe the root cause of the above issue comes from this library.
Long story short, when the timeout parameter is not specified the library uses the default value of 0 when creating SSH connections. This blocked the calling thread, potentially forever (it's still blocked 1 day after we noticed it).
As I mentioned in the Jenkins ticket: