Description of problem: When watching objects such as builds, buildconfigs we have observed that the websocket sometimes close with an abnormal error code (1006 abnormal closure) The websocket will often close within the first few minutes. This makes it challenging to rely on watch/events to trigger external systems and tasks. Some debugging details from the openshift.io Che team: https://github.com/redhat-developer/rh-che/issues/672 We have a reproducer here: https://github.com/redhat-developer/rh-che/issues/672#issuecomment-389808446
I was not able to track where in k8s code we return this to the client. The only trace was: vendor/golang.org/x/net/websocket/hybi.go 34: closeStatusAbnormalClosure = 1006 Which is actually not returned anywhere... I guess this suggests a networking issue between the client and the server? Is there any proxy/loadbalancer between the API server and the client?
initial issue was closed - https://github.com/redhat-developer/rh-che/issues/672#issuecomment-394631744 so I think this can be closed too.