Skip to content

Latest commit

 

History

History
28 lines (20 loc) · 3.53 KB

v1.1.0.md

File metadata and controls

28 lines (20 loc) · 3.53 KB
Raw

HPE CSI Driver for Kubernetes 1.1.0 Release Notes

HPE CSI Driver for Kubernetes 1.1.0

Version: 1.1.0
Revision: Thursday March 12, 2020

Resolved Issues

The following table lists the resolved issues for HPE CSI Driver for Kubernetes v1.1.0.

ID Component Title
CON-597 csi.k8s Bind mounts are left around after pod removal.

Known Issues

The following table lists the known issues for HPE CSI Driver for Kubernetes v1.1.0.

ID Component Title Description Workaround
CON-577 csi.k8s Pods do not move to a new node when kubelet is unavailable In case of node partition, (host unreachable) when the kubelet is unreachable the pods do not move to the new node. This is a known issue with kubernetes kubernetes/kubernetes#65392 1. 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 2. Reboot the failed node
CON-689 csi.k8s After resize operation, new size is not reflected on PVC until its attached to a pod. For devices with mount access type, our driver indicates to external resizer that filesystem resize is required. This causes FileSystemResizePending condition to be added to PVC. This condition is only cleared when PVC is attached to a pod and device/filesystem expansion is done by our node plugin. Only after this, new PVC size will be reflected. More information can be found at https://kubernetes-csi.github.io/docs/volume-expansion.html None
CON-690 csi.k8s Clone of PVC with overridden parameters fails with status 400(BAD_REQUEST) Certain properties like performancePolicy and dedupe cannot be changed on the clone volume to differ from the parent volume. These have to match with the parent volume while creating a clone, both in storage class as well as the PVC overrides. Check that the storage class parameters and the PVC overrides of the clone to match those of the parent volume . These are applicable to properties like performancePolicy (with same block size) and dedupe settings, while others like IOPS can vary.
CON-720 csi.k8s Volumeattachment not cleared immediately after pod deletion This is a known issue with kubernetes where bulk verify of volume attachments is not present. This causes delay to fetch volumeattachments, which gets throttled as mentioned in the issue kubernetes/kubernetes#84169. Workaround as suggested by Saad at kubernetes/kubernetes#84169 (comment) to overcome this issue, i.e set the following in kube-controller-manager.yaml --disable-attach-detach-reconcile-sync=true.
CON-728 csi.k8s Stale mounts left if pod is deleted before nodePublish is completed An issue with external attacher and a race condition where stale mounts (bind mounts) are not cleaned if pod is deleted after nodeStage but before nodePublish is complete kubernetes/kubernetes#89112 None
CON-732 csi.k8s Pod is stuck in creation state for long time after node drain on Ubuntu 18.04 After repeated drain tests, pods can be stuck in ContainerCreating state for long time as node plugin cannot attach and discover device on node. This is due to timeout in multipathd when getting path information. This was seen when there are failed/stale mpath devices/paths on node lying around. Cleanup of stale multipath devices and paths is required to fix timeout by multipathd, or node reboot is required to clear those paths..