Conversation
anjmao
approved these changes
Mar 16, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Background
When a PVC carries the omni.cast.ai/provision-on-all-edge-nodes annotation, every virtual kubelet (VK) should create a corresponding remote PVC on its own cluster. A single local PV is shared across all VKs, with a deterministic name (pvc-), created by whichever VK runs first.
The Bug
Both VKs run the full provisioning flow for provision-on-all-edge-nodes PVCs:
Result: no errors visible on the second VK, but the remote PVC on its cluster was never created.
Fix 1 (controller.go)
When shouldProvisionOnAllEdges=true and PV Create returns AlreadyExists, treat it as success. The other VK created the PV, which is expected — the remote PVC on this cluster was still provisioned successfully.
Fix 2 (controller.go / shouldProvision)
for provision-on-all-edge-nodes PVCs that are already bound (VolumeName set, annSelectedNode gone), check whether the remote PVC exists on this cluster. If not, return true to trigger provisioning. This covers the "missed window" case and also self-heals if a remote PVC is accidentally deleted later.
Fix 3 (remoteprovisioner.go)
buildPvNodeAffinity was calling options.SelectedNode.GetLabels() unconditionally, which panics when SelectedNode is nil — as it is when annSelectedNode is absent (Fix 2 path). Since NodeSelectorOpExists does not use node label values, the label lookup is now scoped to the !shouldProvisionOnAllEdges branch only.