-
Bug
-
Resolution: Obsolete
-
Minor
-
4.11.z
-
None
-
False
-
Description of problem:
A cluster was upgrading from 4.10.45 to 4.11.25 but it got stuck with machine-config operator going down due to one of the machine-config-daemon pods crashing with:
F0206 23:04:46.540997 883811 start.go:154] Failed to initialize single run daemon: error initializing rpm-ostree: error running systemctl start rpm-ostreed: Job for rpm-ostreed.service failed because the control process exited with error code. See "systemctl status rpm-ostreed.service" and "journalctl -xe" for details. : exit status 1
trking also found this from the affected node logs:
$ zgrep -o ' rpm-ostree[[].*' ip-10-44-225-172.ec2.internal.log.gz | sed 's/^[^:]*: //' | sort | uniq -c | sort -n | tail 2 In idle state; will auto-exit in 62 seconds 2 In idle state; will auto-exit in 64 seconds 2 Txn Cleanup on /org/projectatomic/rpmostree1/rhcos successful 3 In idle state; will auto-exit in 63 seconds 3 Locked sysroot 3 Reading config file '/etc/rpm-ostreed.conf' 3 Unlocked sysroot 5 In idle state; will auto-exit in 60 seconds 7 In idle state; will auto-exit in 61 seconds 68 error: The maximum number of active connections for UID 0 has been reached
This is similar to https://bugzilla.redhat.com/show_bug.cgi?id=2111817#c22 however https://github.com/openshift/machine-config-operator/pull/3292/files means that the workaround should have already been applied and we (managed OpenShift SRE) should not have received an alert for the machine config operator going down
Draining the affected worker node did help with the MCD pod being able to start running again.
Version-Release number of selected component (if applicable):
4.11.25 (while upgrading from 4.10.45)