Skip to content

Check for running containers, not ready containers

RLazarus requested to merge work/rzl/running into main

If a mwscript job finishes up very quickly (running Version.php for example) then the httpd container never goes ready -- MediaWiki is already gone by the time the readiness checks start.

Presently, that means the sidecar controller never shuts down the pod: it waits until a container is ready before it considers it to be "running," so as far as the controller is concerned, it's still starting up, forever, even though the primary container is finished.

That's a reasonable default (if a sidecar never went ready, you might want to leave it running to investigate why) but for our purposes, let's consider any container in the "running" state to be running.

Bug: T348284

Merge request reports