Description
Describe the bug
This issue is about two Harvester clusters sharing the same backup storage. This is important when trying to migrate virtual machines between clusters, and for testing that a backed-up virtual-machine is consistent and can be restored successfully. After all, an untested backup is no backup.
Imagine two Harvester clusters sharing one backup storage location. I expect that soon after the backup of a VM has completed on one cluster I should be able to select that backup on the other cluster to restore the VM onto the second cluster. This is not happening.
Looking in the Longhorn user interface in both clusters it can be seen that the backed-up volumes are visible by both clusters. It seems that Harvester is not reading the config files and refreshing its user interface.
In my experience, after a very long wait of maybe 12 hours, a very small number of VMs backed up from the "source" Harvester cluster eventually appear ready for restoration into the "destination" Harvester cluster.
To Reproduce
- In the “Setting: backup-target” screen of two separate Harvester clusters, configure the same shared storage’. In my specific use case the “source” Harvester cluster is v1.3.1 and the “destination” Harvester cluster is v1.4.0 and the storage is NFS.
- In the “Virtual Machines” screen of the “source” Harvester cluster, complete a backup of several VMs
- In the “Backup” tab of the Longhorn user interface of the “source” Harvester cluster, see that the volumes of the VMs are listed.
- In the “Backup” tab of the Longhorn user interface of the “destination” Harvester cluster, after a while see that the volumes of the VMs appear
- In the “Virtual Machine Backups” screen of the “destination” Harvester cluster, note that the backups are not listed
Expected behavior
I expect to see the VMs backed up from one Harvester cluster into storage shared by multiple Harvester clusters to be restorable onto any of the Harvester clusters in a timely manor.
Environment
- Harvester ISO version:
- “source” Harvester cluster is v1.3.1
- “destination” Harvester cluster is v1.4.0
- Underlying Infrastructure: Bare metal nodes and NFS storage.
Additional context
VM Backups screen on "source" Harvester cluster (highliting the only VM that is available in the "destination" Harvester cluster)
VM Backups screen on "destination" Harvester cluster
Backup screen of "source" Longhorne
Backup screen of "destination" Longhorne
Listing of Harvester config files from the NFS storage
Metadata
Metadata
Assignees
Labels
Type
Projects
Status