Workspace stuck in unknown state

My workspace timed out and after pressing start it has now been stuck in an unknown state (for about 20 mins) in the dashboard with a open button which is unresponsive.

how can i get this working again?

Thanks

update:

the state reset itself, can now access my workspace (took over an hour) :slight_smile:

1 Like

Hi @gazzamc,

sorry for the inconvenience. We have a cleanup mechanism that catches workspaces in state UNKNOWN and stops them.

We recently saw more short network issues which we currently do not handle very good. We’re working on improving that.

To expand on @geropl’s reply, if my understanding is correct, here is what happens currently in Gitpod:

  • Occasionally, a short network issue (where Kubernetes randomly says: network is not ready) causes a workspace to briefly get into state UNKNOWN
  • Our workspace manager immediately decides to stop any workspace it sees in state UNKNOWN
  • For some reason, we lose track of that workspace, and so after 1 hour of state UNKNOWN we declare the workspace as stopped, and you can start it again

This is not ideal. Instead, we’re considering the following:

  • Even when there are (brief) network interruptions, we should try to keep workspaces running as-is (unless the owner stops it explicitly – owners should always have control over their workspaces)
  • If a workspace stays in state UNKNOWN for longer than 5 minutes or so, then our workspace manager can stop it (and also the workspace should be marked as stopped so that users don’t have to wait for one hour before starting it again)

But probably the most important thing is to ensure that a final backup happens successfully, because we think that losing data is far worse than waiting for some time.

1 Like

Hitting this issue but the workspace has been in the UNKNOWN state for almost 4 hours now. Is there no way to force the workspace to close?

Lost a decent amount of work unfortunately if so.