Bug 1939259

+++ This bug was initially created as a clone of Bug #1938316 +++

+++ This bug was initially created as a clone of Bug #1936488 +++

test:
[sig-instrumentation][Late] Alerts shouldn't report any alerts in firing state apart from Watchdog and AlertmanagerReceiversNotConfigured 

is failing frequently in CI, see search results:
https://search.ci.openshift.org/?maxAge=168h&context=1&type=bug%2Bjunit&name=&maxMatches=5&maxBytes=20971520&groupBy=job&search=%5C%5Bsig-instrumentation%5C%5D%5C%5BLate%5C%5D+Alerts+shouldn%27t+report+any+alerts+in+firing+state+apart+from+Watchdog+and+AlertmanagerReceiversNotConfigured

Seeing a lot of failures with the queries used.
https://prow.ci.openshift.org/view/gcs/origin-ci-test/logs/release-openshift-origin-installer-e2e-remote-libvirt-image-ecosystem-ppc64le-4.8/1368381121678544896

https://prow.ci.openshift.org/view/gcs/origin-ci-test/logs/periodic-ci-openshift-release-master-nightly-4.8-e2e-gcp-rt/1368886803049746432

Test grids show that these errors have been popping up for a long time
https://testgrid.k8s.io/redhat-openshift-ocp-release-4.8-informing#release-openshift-origin-installer-e2e-remote-libvirt-image-ecosystem-ppc64le-4.8

https://testgrid.k8s.io/redhat-openshift-ocp-release-4.8-informing#periodic-ci-openshift-release-master-nightly-4.8-e2e-gcp-rt

--- Additional comment from Simon Pasquier on 2021-03-08 09:33:51 CST ---

Looking at release-openshift-origin-installer-e2e-remote-libvirt-image-ecosystem-ppc64le-4.8 [1], there's a problem with DNS resolution for Alertmanager pods [2] which leads to "AlertmanagerMembersInconsistent" being fired. It should be redirected to the teams dealing with libvirt and/or ppc64 platforms because it's not something we see in other environments.

Looking at periodic-ci-openshift-release-master-nightly-4.8-e2e-gcp-rt [3], "TargetDown" is firing for the node-exporter and kubelet targets. It means that for some reason, Prometheus fails to scrape metrics and I can see some failures in node_exporter's kube-rbac-proxy [4][5][6] that would match. Again since it is specific to a given job, it would be best redirected to the folks in charge of this job.

[1] https://prow.ci.openshift.org/view/gcs/origin-ci-test/logs/release-openshift-origin-installer-e2e-remote-libvirt-image-ecosystem-ppc64le-4.8/1368381121678544896
[2] https://gcsweb-ci.apps.ci.l2s4.p1.openshiftapps.com/gcs/origin-ci-test/logs/release-openshift-origin-installer-e2e-remote-libvirt-image-ecosystem-ppc64le-4.8/1368381121678544896/artifacts/e2e-remote-libvirt/pods/openshift-monitoring_alertmanager-main-0_alertmanager.log
[3] https://prow.ci.openshift.org/view/gcs/origin-ci-test/logs/periodic-ci-openshift-release-master-nightly-4.8-e2e-gcp-rt/1368886803049746432
[4] https://gcsweb-ci.apps.ci.l2s4.p1.openshiftapps.com/gcs/origin-ci-test/logs/periodic-ci-openshift-release-master-nightly-4.8-e2e-gcp-rt/1367656589833539584/artifacts/e2e-gcp-rt/gather-extra/artifacts/pods/openshift-monitoring_node-exporter-dnw48_kube-rbac-proxy.log
[5] https://gcsweb-ci.apps.ci.l2s4.p1.openshiftapps.com/gcs/origin-ci-test/logs/periodic-ci-openshift-release-master-nightly-4.8-e2e-gcp-rt/1367656589833539584/artifacts/e2e-gcp-rt/gather-extra/artifacts/pods/openshift-monitoring_node-exporter-5jz89_node-exporter.log
[6] https://gcsweb-ci.apps.ci.l2s4.p1.openshiftapps.com/gcs/origin-ci-test/logs/periodic-ci-openshift-release-master-nightly-4.8-e2e-gcp-rt/1367656589833539584/artifacts/e2e-gcp-rt/gather-extra/artifacts/pods/openshift-monitoring_node-exporter-grgw8_kube-rbac-proxy.log

--- Additional comment from Prashanth Sundararaman on 2021-03-09 19:42:22 CST ---

i see this issue on any libvirt deploy on baremetal..not specific to multi-arch, where there is this alert firing:

33.33% of the machine-api-controllers/machine-api-controllers targets in NamespaceNSopenshift-machine-api namespace are down.

this is due to the kube-rbac-proxy-machine-mtrc container reporting these errors:

I0309 17:29:57.278737       1 main.go:159] Reading config file: /etc/kube-rbac-proxy/config-file.yaml
I0309 17:29:57.285997       1 main.go:190] Valid token audiences: 
I0309 17:29:57.286071       1 main.go:278] Reading certificate files
I0309 17:29:57.286336       1 main.go:311] Starting TCP socket on 0.0.0.0:8441
I0309 17:29:57.287013       1 main.go:318] Listening securely on 0.0.0.0:8441
2021/03/09 17:43:04 http: proxy error: dial tcp [::1]:8081: connect: connection refused
2021/03/09 17:43:12 http: proxy error: dial tcp [::1]:8081: connect: connection refused
2021/03/09 17:43:33 http: proxy error: dial tcp [::1]:8081: connect: connection refused
2021/03/09 17:43:42 http: proxy error: dial tcp [::1]:8081: connect: connection refused
2021/03/09 17:44:03 http: proxy error: dial tcp [::1]:8081: connect: connection refused
2021/03/09 17:44:12 http: proxy error: dial tcp [::1]:8081: connect: connection refused
2021/03/09 17:44:33 http: proxy error: dial tcp [::1]:8081: connect: connection refused
2021/03/09 17:44:42 http: proxy error: dial tcp [::1]:8081: connect: connection refused
2021/03/09 17:45:03 http: proxy error: dial tcp [::1]:8081: connect: connection refused

--- Additional comment from Andy McCrae on 2021-03-10 04:13:11 CST ---

This looks to be happening because a change to metrics went in to the machine-api-operator, and the required provider change was not applied to the libvirt provider:

https://github.com/openshift/machine-api-operator/pull/609

This will impact all branches going back to 4.6 - I'm looking into applying the change to the libvirt provider now.

--- Additional comment from Dan Li on 2021-03-11 07:13:45 CST ---

Hi Andy or Prashanth, as part of bug triaging, can we provide a "Severity" for this bug?

--- Additional comment from Andy McCrae on 2021-03-11 10:24:10 CST ---

I've marked this as Medium - This should only impact CI since the libvirt-provider isn't used in supported environments, but none-the-less it impairs out ability to perform accurate CI runs.

We have a PR up for master, but we will need to shepherd this through to 4.6: https://github.com/openshift/cluster-api-provider-libvirt/pull/218

--- Additional comment from Yaakov Selkowitz on 2021-03-12 12:57:50 CST ---

(In reply to Andy McCrae from comment #5)
> We have a PR up for master, but we will need to shepherd this through to
> 4.6: https://github.com/openshift/cluster-api-provider-libvirt/pull/218

Automatic cherry-pick failed, so we'll need manual PRs for 4.7 and 4.6.