daemonset controller should respect taints #31020
Conversation
k8s-github-robot
added
size/M
|
If a daemon pod without a toleration is running on a node and a NoSchedule taint is added, should the daemonset controller kill that pod? Will add tests for this once we decide one way or the other. |
|
GCE e2e build/test passed for commit 00f05b4. |
No, that should only happen for "NoExecute" taint, which is not supported yet. |
| @@ -713,6 +713,16 @@ func (dsc *DaemonSetsController) nodeShouldRunDaemonPod(node *api.Node, ds *exte | |||
| for _, r := range reasons { | |||
| glog.V(2).Infof("GeneralPredicates failed on pod %s for reason: %v", newPod.Name, r.GetReason()) | |||
| } | |||
| if !fit { | |||
| return false | |||
There was a problem hiding this comment.
In GeneralPredicates we accumulate all the failure reasons, so if we want to do the same thing here I guess we'd drop this if statement and continue with checking PodToleratesNodeTaints regardless of whether fit is already false. But since this is only going into a log message I think it's not worth doing it that way (if it were going into events the user could see, then having the full set of failure reasons could be useful).
|
LGTM |
davidopp
added
lgtm
|
@k8s-bot test this [submit-queue is verifying that this PR is safe to merge] |
|
GCE e2e build/test passed for commit 00f05b4. |
|
Automatic merge from submit-queue |
Automatic merge from submit-queue Revert "daemonset controller should respect taints" Reverts #31020 We will be unreverting with some modifications after v1.4. cc @pwittrock @davidopp
cc @dchen1107 @davidopp
This change is