Description of problem: If gaps in metrics are observed by fsync controller they can provide false positives for leader election growth over time series. This can result in sending admin on an invalid path of etcd performance triage.
> Detected leader change increase of 1.25 over 5 minutes on "BareMetal"; disk metrics are: etcd-nfvpe-12.oot.lab.eng.bos.redhat.com=0.0025688888888888875,etcd-nfvpe-13.oot.lab.eng.bos.redhat.com=0.0018824444444444331,etcd-nfvpe-02.oot.lab.eng.bos.redhat.com=0.001912424242424239
The image below shows leader elections(green) with spikes in leader elections(blue) as a result of the query.
This metric query clearly shows the gaps in the collection of metrics from possible networking issues.
Version-Release number of selected component (if applicable):
How reproducible: given condition 100%
Steps to Reproduce:
1. generate gaps in metrics collection by disrupting networking
2. review events from cluster-etcd-operator namespace for Detected leader change increase events.
3. verify etcd is not actually experiencing leader elections `etcd_server_leader_changes_seen_total`
Actual results: leader elections are fasley reported by fsync controller
Expected results: fsync controller should event only with an actual observation of the issue.
This bug verification be blocked by new bug: https://bugzilla.redhat.com/show_bug.cgi?id=2021453
Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.
For information on the advisory (OpenShift Container Platform 4.9.8 bug fix update), and where to find the updated
files, follow the link below.
If the solution does not work for you, open a new bug report.