Workspace hangs on restart after being stopped
- MR: Fix workspaces reconciliation to send inventory... (!187261 - merged) • Vishal Tak • 17.11
- MR for 17.10 backport - Fix workspaces reconciliation to send inventory... (!187679 - merged) • Vishal Tak • 17.10
Description
Workspaces in a stopped state do not restart until after a full reconciliation occurs.
This might be a result of a bug where we omit the secrets that have the inventory annotation but include the configmap with the secret label in a partial reconciliation.
If this is the case, this may cause Kubernetes to delete the secret resources on apply.
-
Current behaviour: Workspace restart after stop hangs till full reconciliation
-
Expected behaviour: Workspace restart does not hang independent of when full reconciliation occurs
-
Steps to reproduce: create a workspace, stop it and attempt to start it again (If you have access to
kubectl, you candescribethe pod to see error of missing secret resource) -
Additional information: please include VSCode version, browser dev tool logs, etc that might be helpful to debug the issue!
Root cause analysis
From Fix workspaces reconciliation to send inventory... (!187261 - merged) • Vishal Tak • 17.11:
This was introduced primarily in Update rails to send terminated workspace info ... (!185872 - merged) . The reason it was not caught in the unit tests is because the specs were incorrect changed here before Update rails to send terminated workspace info ... (!185872 - merged) was merged.
References
Acceptance criteria
-
workspaces do not hang on restart irrespective of full reconciliation interval
Implementation plan
- A fix for this was originally attempted in Fix workspace restart blocked due to missing se... (!186697 - closed) • Safwan Ahmed • 17.11, but the final fix is here: Fix workspaces reconciliation to send inventory... (!187261 - merged) • Vishal Tak • 17.11