🌱 enable leader election of klusterlet-agent on single node managed clusters #727
Conversation
haoqing0110
force-pushed
the
br_agent-leader-election
branch
from
971f439 to
4f59822
Compare
Codecov ReportAttention: Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## main #727 +/- ##
==========================================
+ Coverage 63.48% 63.51% +0.02%
==========================================
Files 185 185
Lines 17827 17838 +11
==========================================
+ Hits 11317 11329 +12
+ Misses 5578 5577 -1
Partials 932 932
Flags with carried forward coverage won't be shown. Click here to find out more. ☔ View full report in Codecov by Sentry. |
pkg/registration/spoke/spokeagent.go
Outdated
| @@ -359,7 +359,10 @@ func (o *SpokeAgentConfig) RunSpokeAgentWithSpokeInformers(ctx context.Context, | |||
| if err := wait.PollUntilContextCancel(bootstrapCtx, 1*time.Second, true, o.internalHubConfigValidFunc); err != nil { | |||
| // TODO need run the bootstrap CSR forever to re-establish the client-cert if it is ever lost. | |||
| stopBootstrap() | |||
| return fmt.Errorf("failed to wait for hub client config for managed cluster to be ready: %w", err) | |||
| // DO NOT return error if context is canceled, allows the leader to release the leadership. | |||
There was a problem hiding this comment.
looks good in general.
Just a question: do we need to doc in which case internalHubConfigValidFunc should return err and it means the agent return with error, and will start a new leader election?
There was a problem hiding this comment.
yes, more comments added.
|
[APPROVALNOTIFIER] This PR is APPROVED This pull-request has been approved by: haoqing0110, qiujian16 The full list of commands accepted by this bot can be found here. The pull request process is described here
Needs approval from an approver in each of these files:
Approvers can indicate their approval by writing |
…sters Signed-off-by: Qing Hao <[email protected]>
haoqing0110
force-pushed
the
br_agent-leader-election
branch
from
4f59822 to
c0f542f
Compare
|
/lgtm |
openshift-merge-bot
bot
merged commit ed367fd
into
open-cluster-management-io:main
Summary
The leader election of the klusterlet-agent is disabled on the single node managed cluster to speed up the restart procedure (open-cluster-management-io/registration-operator#193). While it might cause that two pods with different configuration, such as images or bootstrap config, running at the same time. Some unexpected behavior happens in this situation, for example, when old and new agents both exist, they override hub-kubeconfig-secret with different value and trigger each other create csr constantly.
To fix those issues, it is necessary to enable the leader election of the klusterlet-agent on the single node managed cluster.
Related issue(s)
#695