Will be mounting live volume [] to worker pod [] since it does not support volume snapshots

​@Sougato Roy can you raise this question towards ​@AmitM 

Hi Onno,

Thank you and It was good to hear from you. I'll follow up with you as soon as I have an update.

Thanks,

Roy

Hi Onno

In absence of CSI Snapshot capability on Kubernetes, we mount the application PVC (live volume) to worker pod and read data from it, that's when you see below logging.

503419 7aefc 04/03 23:37:16 23279 K8sApp::AddVolumeToPod() - Will be mounting live volume [nginx-pvc-1] to worker pod [nginx-pvc-1-liv-vol-cv-23279] since it does not support volume snapshots

With CSI, we take a snapshot of the PVC, Create PVC from snapshot and then mount that PVC to worker pod and read from it.

Thanks
Manoranjan

Hi guys,

I was hoping that bK8sSnapFallBackLiveBkp would mute the events but it doesn’t. I also know we don’t have any PVCs as of yet in these clusters so it isn’t even a factor. In fact with bare metal OpenShift there is no CSI - this is all “local” NFS each node provides (ODF I believe).

Do you guys believe that “hiding” the 72:106 events is wise and are you saying that without any CSI in play there is no snapshot capability? Sorry I’m not a K8s guy, just trying to understand whether this is something I need to spend time on with our SME. 

thanks

Snapshot capability comes with CSI, so no CSI implies no snapshot capability.
 

Without snapshots, achieving application-consistent backups is challenging because applications may be writing to the PVC during the backup process, hence the message is logged as a reminder.

 

Yep, understood. But these are etcd and generic namespace backup jobs (for now). So we think ignoring the events is OK?

thanks