This message was deleted.

Hey! I made a post about a similar scenario in the harvester channel and didn't get any traction. I ran through a test scenario where I had a cluster with a number of PVCs backed up by longhorn to S3. I completely destroyed and recreated the cluster using the same IaC configuration, and reconnected longhorn to the backup target. It re-created the volumes from the backup, but I could not find any way to re-bind the PVCs to the specific volumes, as that information was lost. I used a simple container to mount each volume and inspect it, and changed each pod to specifically mount the backed up volume. (This was not a good experience). I imagine you could possibly get around this if, instead of re-creating the cluster from IaC, you restored it's state using the rancher backup restore operator. This should preserve the kube api state, including all those pvc bindings. What would be nice, IMO, is if you could pass some annotation information into the longhorn backup target for the volume so that when it's restored, there's some information about what it was in the past.

Hey, thanks for your comment, I just saw it, I used velero at the end, it works perfectly in conjunction with Longhron via the use of CSI Velero-plugin-for-CSI, in brief, when you create a Velero backup it creates a

VolumeSnapshot

, which tells Longhorn to create a backup. Now, if Longhorn is configured correctly, it writes this backup to a S3 bucket.