Veeam agent for linux and Kubernetes persistant volume

[archvector]

Hi,

I'm currently searching a way to backup a persistant volume (vsphere volume in kubernetes) located on a vsphere VM using Veeam.
Since the persistant volume is seen as an ext4 filesystem at the operating system level, i've installed a veeam agent on the VM and was able to backup/restore that persistant volume successfully by using as source the block device UUID (UUID is persistant)
The main downside of this solution is that i'm forced to provide affinities to the stateful POD that use that persistant volume stick to the VM that have the veeam agent

I was not able to have a clear answer if that solution was supported on not by veeam. Could you please confirm it ?

Thanks!

Notes :
- I know kasten would be the appropriate tool but if i'm not wrong it's not yet possible to send data to VBR using kasten product
- NFS is not possible on my platform
- Binding that particular POD on that VM is not an issue on my platform

[tsightler]

To be clear, you installed VAL on the kerbernetes worker VM, is that correct? What type of backup did you perform, just file level? In general I would say it is not supported, but, based on the exact details, it might work. Were you able to see a snapshot of the volume taken? Were you writing data from the container into this volume at the time of backup? That would be my primary concern, the volume is mapped to the K8S worker node, but it's the container that is actually doing the reading/writing of this volume. In theory, this should be OK, but in practice it is untested and there are many possible corner cases.

[archvector]

Hi tsightler,

- Yes i've installed the VAL on the kubernetes worker VM (which is a RHEL , os supported by VAL)
- I've performed multiple block level backup and i double check all veeam logs in /var/log/veeam. Snapshot worked properly since the block device is formatted in ext4.
- There is no write while the snapshot is done nor while the veeam backup run.
- Yes the persistent volume data is written by some code within the container. I didn't notice a different behavior of the VAL while backing up this filesystem compared to other filesystem mounted on the operating system. Moreover, i believe even the page cache is managed at the kernel level (operating system level)

Do you know if there is one URL or something that is showing it's not supported ? This would help me to close this topic with my customer

Thanks,

[tsightler]

I don't believe there is any official URL that list everything that is not supported, rather we explicitly list things that are supported and K8S with persistent volumes is not on that list. They way you have it setup I believe it would be OK as long as the volume is always presented to that host and never on any other hosts, but if someone called expecting support for the K8S portions, for example, how to reclaim the volume if it were lost, etc. then I don't think support could help with those things as that's beyond the function of VAL.

[archvector]

Ok, thanks