Open skinowski opened 6 years ago
Second instance of this again. Looks like same tests; TestRouteRunnerIOPipes at first invocation of /r/zoo/http.
From docker src code, seems like two ways this pause can already be set, events from containerd or dockerd pause itself.
it doesn't seem like it would be horrid to ignore any 'already paused' errors and just leave it paused, since we are trying to pause. due to retries, 'docker start' we are already doing this with 'container already running' errors, that could be one place this comes from but i don't see any 'docker temporary error' logs to indicate that. seems like docker pause command could return before container is actually paused? :/
I agree, but right now we do recover from this by discarding the hot container, so it's not visible to callers for the moment... Just tracking it here to make sure we eventually find out the root cause.
Sample snippet from test logs:
For this container, the following ops are performed:
From fn perspective, things look right, but last operation fails with "already paused". This is either a conflict or a docker issue. Or something just squezeed in between pausing it before fn server. Notice that last request was just run successfully so the container might have not been paused at all.