leader election bugfix: Delete evicted leader pods #2210
Merged
Conversation
This file contains 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
Before this patch, when the leader pod is hard evicted but not deleted the leader lock configmap is not garbage collected and subsequent operator pods can never become leader. With this patch, an operator attempting to become the leader is able to delete evicted leader pods triggering garbage collection and allowing leader election to continue. Sometimes, evicted operator pods will remain, even with this patch. This occurs when the leader operator pod is evicted and a new operator pod is created on the same node. In this case, the new pod will also be evicted. When an operator pod is created on a non-failing node, leader election will delete only the evicted leader pod, leaving any evicted operator pods that were not the leader. To replicate the evicted state, I used a `kind` cluster with 2 worker nodes with altered kubelet configuration and a memory-hog version of the memcached operator. See the [altered operator docs](https://github.com/asmacdo/go-memcahced-operator/blob/explosive-operator/README.md)
openshift-ci-robot
added
the
size/M
asmacdo
commented
Nov 18, 2019
openshift-ci-robot
added
size/S
joelanford
reviewed
Nov 18, 2019
openshift-ci-robot
added
size/M
joelanford
reviewed
Nov 20, 2019
|
When is this going to make it into a release? I see this was merged 20 days ago, but 5 days ago 0.12 was released and this change was not included: https://github.com/operator-framework/operator-sdk/blob/v0.12.0/pkg/leader/leader.go#L112 |
|
Hello! Did this one get into 0.13? I can't see in the CHANGELOG. We are hitting this bug quite frequently lately at work, so it would be of great help to us if this could gets released. |
|
@jmazzitelli @poros this got into v0.13.0. We missed a changelog entry, will follow up with one. |
|
Thanks! :) |
estroz
added a commit
to estroz/operator-sdk
that referenced
this pull request
Dec 12, 2019
estroz
pushed a commit
that referenced
this pull request
Dec 12, 2019
estroz
added a commit
to estroz/operator-sdk
that referenced
this pull request
Dec 12, 2019
estroz
pushed a commit
that referenced
this pull request
Dec 19, 2019
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.
Description
Before this patch, when the leader pod is hard evicted but not deleted
the leader lock configmap is not garbage collected and subsequent
operator pods can never become leader. With this patch, an operator
attempting to become the leader is able to delete evicted leader pods
triggering garbage collection and allowing leader election to continue.
Replication
To replicate the evicted state, I used a
kindcluster with 2 workernodes with altered kubelet configuration and a memory-hog version of the
memcached operator.
See the replication readme
FYI
Sometimes, evicted operator pods will remain, even with this patch.
This occurs when the leader operator pod is evicted and a new operator
pod is created on the same node. In this case, the new pod will also be
evicted. When an operator pod is created on a non-failing node, leader
election will delete only the evicted leader pod, leaving any evicted
operator pods that were not the leader.
Closes #1305
Closes #1874