Red Hat OpenShift Data Foundation storage node failure

You can do a node replacement proactively for an operational node and reactively for a failed node. For a failed node backed by local storage devices, you must replace the Red Hat® OpenShift® Data Foundation storage node.

Before you begin: Red Hat recommends that replacement nodes are configured with similar infrastructure, resources, and disks as the node planned for replacement.
Note: Contact IBM Support before you proceed with any of these fixes.

Procedure

Do the following steps to check for the occurrence of Red Hat OpenShift Data Foundation storage node failure and identify the failed node:

Set the Red Hat OpenShift Data Foundation cluster to maintenance mode:

oc label odfclusters.odf.isf.ibm.com -n ibm-spectrum-fusion-ns odfcluster "odf.isf.ibm.com/maintenanceMode=true"

Example output:

[root@fu40 ~]# oc label odfclusters.odf.isf.ibm.com -n ibm-spectrum-fusion-ns odfcluster "odf.isf.ibm.com/maintenanceMode=true"
odfcluster.odf.isf.ibm.com/odfcluster labeled

Identify the failed node:

Log in to IBM Fusion user interface.
Go to Data foundation page and check for warning in the Health section for storage cluster.

Alternatively, you can use the oc command to identify the node:

oc get node -l cluster.ocs.openshift.io/openshift-storage=

Sample output:

[root@fu71-f09-vm3 ~]# oc get node -l cluster.ocs.openshift.io/openshift-storage=
NAME                               STATUS     ROLES    AGE   VERSION
f09-prc4m-worker-cluster-b-9chb5   NotReady   worker   27d   v1.24.0+4f0dd4d
f09-prc4m-worker-cluster-c-mfb77   Ready      worker   31d   v1.24.0+4f0dd4d
f09-prc4m-worker-cluster-d-r5bxx   Ready      worker   27d   v1.24.0+4f0dd4d

Identify the failed mon (if any) and Red Hat OpenShift Dedicated pods that are running in the node, which is planned for replacement:
In an operational storage node environment:
```
oc get pods -n openshift-storage -o wide | grep -i <node_name>
```

If the storage node failed in a failed storage node, there is no node_name for the failed pods. Filter the pending pods instead.

oc get pods -n openshift-storage -o wide | grep -i pending

Example output: The mon deployment is rook-ceph-mon-d, and the Red Hat OpenShift Dedicated deployment is ook-ceph-osd-0.

[root@fu71-f09-vm3 ~]# oc get pods -n openshift-storage -o wide | grep -i pending
 rook-ceph-mon-d-67686857d7-zv62c                                  0/2     Pending     0          8m50s   <none>            <none>                             <none>           <none>
 rook-ceph-osd-0-75b954c9bf-62xm4                                  0/2     Pending     0          8m50s   <none>            <none>                             <none>           <none>

Remove the failed objects.

Remove the failed node from odfcluster CR

spec:
  autoScaleUp: false
  creator: CreatedByFusion
  deviceSets:
  - capacity: "0"
    count: 3
    name: ocs-deviceset-ibm-spectrum-fusion-local
    storageClass: ibm-spectrum-fusion-local
  encryption:
    keyManagementService: {}
  localVolumeSetSpec:
    deviceTypes:
    - disk
    - part
    size: 2Ti
  storageNodes:
  - f09-prc4m-worker-cluster-d-r5bxx
  - f09-prc4m-worker-cluster-c-mfb77
  - f09-prc4m-worker-cluster-b-9chb5 <<-- this one, remove this line.

Remove the mon and Red Hat OpenShift Dedicated pods

Scale down the deployments of the identified pods. The mon deployment is rook-ceph-mon-d and the Red Hat OpenShift Dedicated deployment is rook-ceph-osd-0.

oc scale deployment rook-ceph-mon-d --replicas=0 -n openshift-storage
oc scale deployment rook-ceph-osd-0 --replicas=0 -n openshift-storage

Ensure that you confirm the values of mon_id and osd_id.

Remove the crashcollector pods

Remove the crashcollector pods ( if any). You must put scale replica to 0.

oc scale deployment --selector=app=rook-ceph-crashcollector,node_name=<node_name>  --replicas=0 -n openshift-storage

Mark the failed node as unschedulable

Mark the node as SchedulingDisabled.

oc adm cordon <node_name>

Example command and output:

oc adm cordon f09-prc4m-worker-cluster-b-9chb5
node/f09-prc4m-worker-cluster-b-9chb5 cordoned

oc get node -l cluster.ocs.openshift.io/openshift-storage=

NAME                               STATUS                        ROLES    AGE   VERSION
f09-prc4m-worker-cluster-b-9chb5   NotReady,SchedulingDisabled   worker   28d   v1.24.0+4f0dd4d
f09-prc4m-worker-cluster-c-mfb77   Ready                         worker   31d   v1.24.0+4f0dd4d
f09-prc4m-worker-cluster-d-r5bxx   Ready                         worker   27d   v1.24.0+4f0dd4d

Remove the pods which are in Terminating state

This step is for the failed storage node. You can ignore this step if your are removing an operational node.

oc get pods -A -o wide | grep -i <node_name> |  awk '{if ($4 == "Terminating") system ("oc -n " $1 " delete pods " $2  " --grace-period=0 " " --force ")}'

Example command and output:

oc get pods -A -o wide | grep -i f09-prc4m-worker-cluster-b-9chb5 |  awk '{if ($4 == "Terminating") system ("oc -n " $1 " delete pods " $2  " --grace-period=0 " " --force ")}'

warning: Immediate deletion does not wait for confirmation that the running resource has been terminated. The resource may continue to run on the cluster indefinitely.
pod "isf-data-protection-operator-controller-manager-5c7cf574d5ms4xx" force deleted

Drain the node

oc adm drain <node_name> --force --delete-emptydir-data=true --ignore-daemonsets

Example command and output:

oc adm drain f09-prc4m-worker-cluster-b-9chb5 --force --delete-emptydir-data=true --ignore-daemonsets
node/f09-prc4m-worker-cluster-b-9chb5 already cordoned

WARNING: ignoring DaemonSet-managed Pods: openshift-cluster-csi-drivers/vmware-vsphere-csi-driver-node-7696f, openshift-cluster-node-tuning-operator/tuned-fk949, openshift-dns/dns-default-gvv4m, openshift-dns/node-resolver-t5dk8, openshift-image-registry/node-ca-wtnp9, openshift-ingress-canary/ingress-canary-kgxts, openshift-local-storage/diskmaker-discovery-qkmh7, openshift-local-storage/diskmaker-manager-m9q42, openshift-machine-config-operator/machine-config-daemon-252j8, openshift-monitoring/node-exporter-cghwc, openshift-multus/multus-additional-cni-plugins-mkz4m, openshift-multus/multus-bz789, openshift-multus/network-metrics-daemon-57v5r, openshift-network-diagnostics/network-check-target-n6bhw, openshift-sdn/sdn-fhp47, openshift-storage/csi-cephfsplugin-5vsp9, openshift-storage/csi-rbdplugin-bfpfs
node/f09-prc4m-worker-cluster-b-9chb5 drained

Delete the node

Delete the failed node:

oc delete node <node_name>

If you do not want to destroy this node for test purpose, you can remove the storage label cluster.ocs.openshift.io/openshift-storage=.

oc label nodes/f09-prc4m-worker-cluster-b-9chb5 cluster.ocs.openshift.io/openshift-storage-

Add new OpenShift compute nodes.

Create a new compute node and ensure that the new node is in ready state.

Update new node info in odfcluster CR.

Edit the odfcluster cr and add new node name in StorageNodes

oc edit odfclusters.odf.isf.ibm.com -n ibm-spectrum-fusion-ns odfcluster

storageNodes:
    - f09-prc4m-worker-cluster-b-djplp  <<--- new node
    - f09-prc4m-worker-cluster-d-r5bxx
    - f09-prc4m-worker-cluster-c-mfb77

Verify whether the new nodes are labeled successfully as storage node.

oc get node -l cluster.ocs.openshift.io/openshift-storage

NAME                               STATUS   ROLES    AGE   VERSION
f09-prc4m-worker-cluster-b-djplp   Ready    worker   27d   v1.24.0+4f0dd4d <<-- this node
f09-prc4m-worker-cluster-c-mfb77   Ready    worker   31d   v1.24.0+4f0dd4d
f09-prc4m-worker-cluster-d-r5bxx   Ready    worker   27d   v1.24.0+4f0dd4d

Verify whether the new PVs are created automatically. The local PV gets created automatically in a short time.

oc get pv | grep Available

local-pv-e97b23d7                          2Ti        RWO            Delete           Available                                                                             ibm-spectrum-fusion-local              2m49s

Replace the failed Red Hat OpenShift Dedicated disks.

Remove the failed Red Hat OpenShift Dedicated from the cluster. You can also specify multiple failed ODs. Use the correct failed_osd_id.

The failed_osd_id is the integer in the pod name immediately after the rook-ceph-osd prefix. You can add comma separated Red Hat OpenShift Dedicated IDs in the command to remove more than one Red Hat OpenShift Dedicated, for example, FAILED_OSD_IDS=0,1,2.

Remove the failed Red Hat OpenShift Dedicated:


oc process -n openshift-storage ocs-osd-removal \
-p FAILED_OSD_IDS=<failed_osd_id> FORCE_OSD_REMOVAL=true | oc create -n openshift-storage -f -

Example output:


[root@fu71-f09-vm3 ~]# oc process -n openshift-storage ocs-osd-removal -p FAILED_OSD_IDS=0 FORCE_OSD_REMOVAL=true | oc create -n openshift-storage -f -
Warning: would violate PodSecurity "restricted:latest": allowPrivilegeEscalation != false (container "operator" must set securityContext.allowPrivilegeEscalation=false), unrestricted capabilities (container "operator" must set securityContext.capabilities.drop=["ALL"]), runAsNonRoot != true (pod or container "operator" must set securityContext.runAsNonRoot=true), seccompProfile (pod or container "operator" must set securityContext.seccompProfile.type to "RuntimeDefault" or "Localhost")
job.batch/ocs-osd-removal-job created

Check the status of the ocs-osd-removal-job pod to verify whether the Red Hat OpenShift Dedicated got removed successfully. A status of Completed confirms that the Red Hat OpenShift Dedicated removal job succeeded.

oc get pod -l job-name=ocs-osd-removal-job -n openshift-storage

Example output:

[root@fu71-f09-vm3 ~]# oc get pod -n openshift-storage  | grep ocs-osd-removal
ocs-osd-removal-job-ls65l                                         0/1     Completed   0          23s

Ensure that the Red Hat OpenShift Dedicated removal is completed:

oc logs -l job-name=ocs-osd-removal-job -n openshift-storage --tail=-1 | egrep -i 'completed removal'

Example output:

[root@fu71-f09-vm3 ~]# oc logs -l job-name=ocs-osd-removal-job -n openshift-storage --tail=-1 | egrep -i 'completed removal'
2022-11-24 15:19:28.750910 I | cephosd: completed removal of OSD 0

Delete the ocs-osd-removal-job:

oc delete -n openshift-storage job ocs-osd-removal-job

Delete the Released PV which is attached to the previous node.

oc get pv | grep -i released


local-pv-e4a12175                          2Ti        RWO            Delete           Released    openshift-storage/ocs-deviceset-ibm-spectrum-fusion-local-0-data-1m6gnz   ibm-spectrum-fusion-local              3h34m
[root@fu71-f09-vm3 ~]# oc delete pv local-pv-e4a12175
persistentvolume "local-pv-e4a12175" deleted

Recover the failed objects.

Restart the mon deployment/pod:

Update the nodeSelector in deployment with new node.

oc edit deployment -n openshift-storage rook-ceph-mon-d

nodeSelector:
        kubernetes.io/hostname: f09-prc4m-worker-cluster-b-djplp <<--new node

Scale the replica to 1 and wait till the mon pods are in running state.

oc scale deployment rook-ceph-mon-d --replicas=1 -n openshift-storage

Example:

oc scale deployment rook-ceph-mon-d --replicas=1 -n openshift-storage
    deployment.apps/rook-ceph-mon-d scaled


[root@fu71-f09-vm3 ~]# oc get pod -n openshift-storage | grep mon
    rook-ceph-mon-a-5bbb9dd98b-z54fx                                  2/2     Running     0          4m45s
    rook-ceph-mon-b-7fdd8f958b-lk9g2                                  2/2     Running     0          5m18s
    rook-ceph-mon-d-6945fbbfc5-nhhw8                                  2/2     Running     0          5m41s

Verify the Red Hat OpenShift Dedicated pods.

Wait till all the pods are in running state.

oc get pods -o wide -n openshift-storage| grep osd

[root@fu71-f09-vm3 ~]# oc get pods -o wide -n openshift-storage| grep osd
rook-ceph-osd-0-d559cc4fb-xspr8                                   2/2     Running     0          4m58s   10.129.6.49       f09-prc4m-worker-cluster-b-djplp   <none>           <none>
rook-ceph-osd-1-6df7f9c669-n94md                                  2/2     Running     0          5m20s   10.128.6.95       f09-prc4m-worker-cluster-d-r5bxx   <none>           <none>
rook-ceph-osd-2-5c5d48ff7c-sdd7l                                  2/2     Running     0          5m17s   10.129.4.125      f09-prc4m-worker-cluster-c-mfb77   <none>           <none>
rook-ceph-osd-prepare-1bf7dd3d71fe899383e625dd0c27ea37-x9vtk      0/1     Completed   0          4h8m    10.128.6.79       f09-prc4m-worker-cluster-d-r5bxx   <none>           <none>
rook-ceph-osd-prepare-24272b5641dc95baffc7932d78894e3c-zhz8m      0/1     Completed   0          5m24s   10.129.6.48       f09-prc4m-worker-cluster-b-djplp   <none>           <none>
rook-ceph-osd-prepare-6f3c3b4626ec9888b6fbe5597afd55ea-zh7cp      0/1     Completed   0          4h8m    10.129.4.113      f09-prc4m-worker-cluster-c-mfb77   <none>           <none>

Verify the Red Hat OpenShift Dedicated encryption settings. If cluster wide encryption is enabled, make sure the “crypt” keyword beside the ocs-deviceset name(s)

oc debug node/<new-node-name> -- chroot /host dmsetup ls

Example output:

# oc debug node/fu47 -- chroot /host dmsetup ls
Starting pod/fu47-debug ...
To use host binaries, run `chroot /host`
ocs-deviceset-sc-lvs-0-data-0clwxf-block-dmcrypt	(253:0)

If verification fails, contact IBM support .

Exit maintenance mode after all steps are completed.

oc label odfclusters.odf.isf.ibm.com -n ibm-spectrum-fusion-ns odfcluster "odf.isf.ibm.com/maintenanceMode-"

Example output:

[root@fu40 ~]# oc label odfclusters.odf.isf.ibm.com -n ibm-spectrum-fusion-ns odfcluster "odf.isf.ibm.com/maintenanceMode-"
odfcluster.odf.isf.ibm.com/odfcluster unlabeled

Go to Data foundation page in IBM Fusion user interface and check the health of the Storage cluster in the Health section.