Fixes Restore for Workerless Clusters #9265
Conversation
- helps restore on workerless clusters - logs error message, but continues Signed-off-by: Nico Weisenauer <nico.weisenauer@sap.com>
Signed-off-by: Nico Weisenauer <nico.weisenauer@sap.com>
Signed-off-by: Nico Weisenauer <nico.weisenauer@sap.com>
|
IMO, this fix MAY fix your problem, but it delays the error in normal use cases. I'm a little reluctant to make it upstream. Because in general we don't support velero running on another cluster at this moment and this looks like a workaround and is very limited. |
reasonerjt
added
the
Needs triage
|
Any other opinions on this? |
| if err := crClient.Get(context.TODO(), types.NamespacedName{Name: veleroDeploymentName, Namespace: namespace}, deployment); err != nil { | ||
| logger.WithError(err).Warnf("Could not find deployment %q in namespace %q", veleroDeploymentName, namespace) | ||
| } else { | ||
| image = veleroutil.GetVeleroServerImage(deployment) |
There was a problem hiding this comment.
So based on feedback saw earlier, my tip for getting this in would be to hide this behavior behind a server feature flag or something.
Like velero server --workerless-mode and you will then perhaps not have to use warn, but just info as it is an expected operating mode.
There was a problem hiding this comment.
Does that sound reasonable? @reasonerjt perhaps you want this implementation explained and detailed as a design for the epic #7843
3 participants
Please add a summary of your change
Fixes restore for workerless clusters. Previously restore of custom resources would crash, because Velero Deployment is not running in the workerless cluster. This simple change prevents the crash and does not change behaviour for "normal" usage with standard clusters.
Does your change fix a particular issue?
Fixes #9251
Please indicate you've done the following:
make new-changelog) or comment/kind changelog-not-requiredon this PR.site/content/docs/main. (not needed imo)