| CON-455 |
csi.k8s |
CSI installation using Helm fails if CRD exists |
If there is an existing CRD (Custom Resource Definition), then an installation using Helm Chart fails. This issue has been seen with the 2.x version of Helm. |
To bypass the CRD installation, include the --no-hooks option to the Helm installation. |
| CON-488 |
csi.k8s |
allowOverrides parameter limitations for storage class during clone or importVolume oprations |
When you perform a volume clone or import operation from the storage class for PersistentVolumeClaim (PVC) using the allowOverrides parameter, some of the attributes are not overridden. These include the size and performancePolicy attributes. This happens because NimbleOS restricts some changes that affect the behavior of a volume. |
If a parameter cannot be overridden, the kubectl describe pvc <PVC_Name> command along with the nimble-csp.log file report error messages. You should modify the PVC spec to remove the override. |
| CON-577 |
csi.k8s |
Pods do not move to a new node when kubelet is unavailable |
There is a known issue with Kubernetes where a pod does not move to a new node if the kubelet in that new node is down.For more information, see kubernetes/kubernetes#65392 |
You can force a move to another node after the NodeEviction Timeout has expired by entering the following command: kubectl delete pod <Pod_Name> --force --grace-period=0 |
| CON-597 |
csi.k8s |
Bind mounts remain after pod is removed |
Bind mounts sometimes remain after a pod cleanup completes. Having these bind mounts left behind can eventually cause the device cleanup to fail, which leaves behind device and iSCSI sessions. |
If there are iSCSI sessions as well as stale bind mounts, you must log out of the target using the command iscsiadm -m node -u <Iscsi_Target>. Doing this will allow the volume to be connected to other hosts. |
| CON-630 |
csi.k8s |
Structure needs cleaning error reported |
This error occurs when a file system is being created on a heavily loaded system that has CPU usage at a maximum. In this environment, the mkfs command takes a long time to complete, which causes the command to time out. This results in the creation of an incomplete file system. |
Reduce the workload on the system to lower CPU usage or distribute the workload so that the mkfs command can complete. |
| CON-637 |
csi.k8s |
Volume name changed on array if size mismatch with PVC Spec found during import |
When an existing volume is imported to the array, the volume name is changed to the new PersistentVolume(PV) name. If there is a size mismatch between the actual volume size and what is specified in the PersistentVolumeClaim (PVC) Spec, the PV create operation fails with the following error message:Volume with size x exists, but a different size y was requested.This leaves the import volume changed to the new spec as per the PVC. |
The size of the import volume must match the size specified by the PVC Spec to prevent this error. If you have already experienced this error, use the new volume name shown in the error message for the import operation. |