Description of problem: Updating the storage.spec.logLevel doesn't take effective, the vmware-vsphere-csi-driver-operator pod doesn't restart and doesn't run as expected logLevel Version-Release number of selected component (if applicable): 4.8.0-0.nightly-2021-04-09-222447 How reproducible: Always Steps to Reproduce: 1. Set Up an OCP4.8 cluster on vSphere 2. Enable TechPreviewNoUpgrade featureset 3. Edit the logLevel for storage.cluster to TraceAll $ oc get storage cluster -o json | jq .spec { "logLevel": "TraceAll", "managementState": "Managed", "operatorLogLevel": "Debug" } 4. Check vmware-vsphere-csi-driver-operator pod Actual results: The vmware-vsphere-csi-driver-operator pod doesn't not restart And the clustercsidriver.spec.operatorLogLevel is "Normal" $ oc get clustercsidriver csi.vsphere.vmware.com -o json | jq .spec.operatorLogLevel "Normal" Expected results: The vmware-vsphere-csi-driver-operator pod should restart and run as TraceAll log level
After changing the storage.spec.logLevel to "TraceAll", the vsphere-csi-driver-operator did not run with TraceAll logLevel. Change status to "Assigned" back. $ oc get storage cluster -o json | jq .spec { "logLevel": "TraceAll", "managementState": "Managed", "operatorLogLevel": "Normal" } $ oc get clustercsidriver csi.vsphere.vmware.com -o json | jq .spec.operatorLogLevel "Normal" $ oc -n openshift-cluster-csi-drivers rsh vmware-vsphere-csi-driver-operator-55c8ddfdb7-c87pv sh-4.4$ ps -aux USER PID %CPU %MEM VSZ RSS TTY STAT START TIME COMMAND 1000270+ 1 0.7 0.4 1742704 80348 ? Ssl 10:44 0:04 /usr/bin/vmware-vsphere-csi-driver-operator start 1000270+ 20 0.1 0.0 12028 3192 pts/0 Ss 10:52 0:00 /bin/sh 1000270+ 26 0.0 0.0 47548 3468 pts/0 R+ 10:52 0:00 ps -aux
@Wei, since PR 131 [1] we no longer cascade the log level from storage CR down to clustercsidriver CR. CSO only creates the clustercsidriver CR if it doesn't exist, but it doesn't change it once it has been created. If the user wants to edit the clustercsidriver CR (to change the log level, for instance), they have to change the clustercsidriver directly. This change was done to accommodate the CSI Config Observer Controller into CSO (as stated in the PR description), but we'll also need this behavior to allow users to configure CSI operators individually (for opting-out of features or removing the operators completely, for example). https://github.com/openshift/cluster-storage-operator/pull/131
Hi Fabio, Yes, we tested that if update storage.spec.logLevel -> the csi driver operator pod log level changed, but clustercsidriver.spec.operatorLoglevel NOT changes. We are okay to the change and close this bz.
Closing as per comments #3 and #4.