rbd-nbd process was killed when upgrade csi-rbdplugin pod

[qiankai-Kwai]

Describe the bug

I use rbd-nbd mounter in storageclass parameters, but I found rbd-nbd process was killed when upgrade csi-rbdplugin daemonset. This caused all rbd block devices dit ot mapped to pod, all pods not work normally.

I want to know how to prevent rbd-nbd process killed when upgrace csi-rbdplugin? Thanks a lot.

[humblec]

unfortunately known limitation on rbd nbd driver @qiankai-Kwai https://github.com/ceph/ceph-csi/issues/667

[qiankai-Kwai]

unfortunately known limitation on rbd nbd driver @qiankai-Kwai #667

Any plans to solve this problem in the future?

[qiankai-Kwai]

we will not resolve this limitation?

[Madhu-1]

@qiankai-Kwai we are working on a new design for the rbd-nbd and cephfs fuse clients. we don't have any deadline for this one yet. @pkalever can update more on this one.

[humblec]

we will not resolve this limitation?

@qiankai-Kwai check on this PR too https://github.com/ceph/ceph-csi/pull/1839/, especially the options like try-netlink, reattach.. params

[pkalever]

Yes, PR https://github.com/ceph/ceph-csi/pull/1839 is to showcase the POC model of working with rbd-nbd at ceph-csi

We are currently working on the design to integrate the rbd-nbd support at ceph-csi, you can soon expect an upstream design doc (say in a couple of weeks from now) and we still might take a couple of releases to add the complete support to our ceph-csi code base.

[stale[bot]]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed in a week if no further activity occurs. Thank you for your contributions.

[pkalever]

@Madhu-1 @nixpanic given we have volume healer now, can we move this to closed, please?

[Madhu-1]

Yes closing this one, Thanks @pkalever for fixing it :+1: