Closed mboersma closed 7 years ago
PR issued in https://github.com/kubernetes/charts/pull/982; chart installed on staging cluster for testing and docs updated with this info.
Oops; accidentally closed. TODOs remaining for this issue are:
workflow-e2e
chartspotify-docker-gc
chart is installed on any leasable e2e clustersThis needs to wait until the chart is in stable; see https://github.com/kubernetes/charts/pull/982
Work completed in https://github.com/deis/workflow-e2e/pull/362 ; closing.
The e2e chart currently uses a DaemonSet with the spotify-gc container to clean up stale Docker image layers and keep the graph from getting too large. But this is clumsy; it continually fails and
CrashLoop
s with an exponential backoff that becomes stupidly large very quickly.There must be some way to use the new CronJob construct to run a garbage collector sidecar on each node with a regular schedule.