d8dfun/virtual-kubelet
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases 1 Wiki Activity

Allow providers to update pod statuses

Browse Source 
We had added an optimization that made it so we dedupe pod status updates
from the provider. This ignored two subfields that could be updated along
with status.

Because the details of subresource updating is a bit API server centric,
I wrote an envtest which checks for this behaviour.

Signed-off-by: Sargun Dhillon <sargun@sargun.me>
This commit is contained in:
Sargun Dhillon
2021-02-15 17:34:01 -08:00
parent 7feb175720
commit c4582ccfbc
4 changed files with 89 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
node/podcontroller.go
View File

@@ -339,7 +339,11 @@ func (pc *PodController) Run(ctx context.Context, podSyncWorkers int) (retErr er
kPod := obj.(*knownPod)
kPod.Lock()
if kPod.lastPodStatusUpdateSkipped && !cmp.Equal(newPod.Status, kPod.lastPodStatusReceivedFromProvider.Status) {
if kPod.lastPodStatusUpdateSkipped &&
(!cmp.Equal(newPod.Status, kPod.lastPodStatusReceivedFromProvider.Status) ||
!cmp.Equal(newPod.Annotations, kPod.lastPodStatusReceivedFromProvider.Annotations) ||
!cmp.Equal(newPod.Labels, kPod.lastPodStatusReceivedFromProvider.Labels) ||
!cmp.Equal(newPod.Finalizers, kPod.lastPodStatusReceivedFromProvider.Finalizers)) {
// The last pod from the provider -> kube api server was skipped, but we see they no longer match.
// This means that the pod in API server was changed by someone else [this can be okay], but we skipped
// a status update on our side because we compared the status received from the provider to the status