Description of problem: ~~~From Customer Description~~~ Fluentd creates /etc/hostname directory and break systemd-hostnamed.service on the nodes Basically /etc/hostname does not exist on our nodes (they are purely dhcp driven) and when fluentd pod starts, it creates /etc/hostname path as a directory, due to following piece in daemon set: - name: dockerhostname hostPath: path: /etc/hostname type: '' /etc/hostname being a directory causes systemd-hostnamed.service to fail at the subsequent boot with: Nov 25 13:02:56 localhost systemd-hostnamed[1859]: Failed to read hostname and machine information: Is a directory Nov 25 13:02:56 localhost systemd[1]: Started Hostname Service. This causes the node to revert its hostname to 'localhost' - and basically self-destruct. According to this 1746968 bug, this is allegedly fixed in 4.2 - but it is not clear what the actual fix is? Has the daemon set template been changed in any way? Is /etc/hostname supposed to exist as a file in every case? Regardless of the actual cause, we are definitively still experiencing this problem on 4.2 - and this is brand new 4.2 deployment, not an upgrade from 4.1.x (which - if it was the case - could maybe explain existence of some stale, buggy deployment) ~~~ Version-Release number of selected component (if applicable): OCP 4.2.7 with Logging Operator 4.2.5 How reproducible: Customer verified easily and repeatedly reproducible Additional info: Logging dump data and must-gather in ticket (too large to attach to bz)
*** This bug has been marked as a duplicate of bug 1746968 ***
(In reply to Greg Rodriguez II from comment #3) > Please see the detail from the description that is pertinent to c#2: > > ~~~ > > According to this 1746968 bug, this is allegedly fixed in 4.2 - but it is > not clear what the actual fix is? The fix is to remove the mount point [1] as shown that it does NOT EXIST in 4.2 [2] [1] https://github.com/openshift/cluster-logging-operator/pull/239/files#diff-977fde508b09ae1d524afe3125ac12f1L285 [2] https://github.com/openshift/cluster-logging-operator/blob/release-4.2/pkg/k8shandler/fluentd.go#L280-L289 > Has the daemon set template been changed in any way? Is /etc/hostname > supposed to exist as a file in every case? See [1] > > Regardless of the actual cause, we are definitively still experiencing this > problem on 4.2 - and this is brand new 4.2 deployment, not an upgrade from > 4.1.x (which - if it was the case - could maybe explain existence of some > stale, buggy deployment) This is only possible if there is a bug in the operator where it is not updating the DS. You can look to see for certain but the code I pointed you to is evidence that fix is in the 4.2 branch. Alternatively are you certain you have properly upgraded logging? The initial 4.1 release subscriptions pointed to the 'preview' channel where subsequent releases are associated with a '4.x' channel. > > ~~~ > > I have acknowledged that 1746968 exists and is closed per errata, however I > have a specific question. > > Please reopen ticket as it is not resolved.