Snapshot from CephFS StorageClasse consumes 1:1 data size?

[kgts23]

Hello,

I am using ceph-csi in my k8s cluster.

I am trying to backup some of my PVCs, one of the PVs being created with CephFS (RWX mode is required). This PV has a size of 5.1 TB. When I take a snapshot (using veeam-k10), I can see that the size of the snapshot is exactly the same size as the source PV. Therefore, my Ceph storage capacity is exhausted.

Is a snapshot really that expensive, or is there something wrong? I'm wondering because I found the design proposal where this problem has already been discussed and implemented (https://github.com/ceph/ceph-csi/blob/devel/docs/design/proposals/cephfs-snapshot-shallow-ro-vol.md). Or am I wrong?

Best regards

[Madhu-1]

@kgts23 snapshot doesn't consume any storage unless they are restored to a new RW PVC, where do you see this consumption? in the volumesnapshot object? we don't have snapshot size in snapshot info in ceph, we are returning the parent subvolume size for the snapshot size in csi.

[kgts23]

Hey @Madhu-1,

While creating a snapshot in k8s trough k10, I see how the storage space in Ceph is continuously decreasing (ceph -s). My placement groups are replicated with three copies. I can therefore observe that after several hours with a net storage consumption of 5.1 TB, the CEPH storage has grown by exactly 15.3 TB gross storage consumption.

What I haven't checked yet is whether k10 binds a RW-PVC to consolidate/verify the snapshot contents.

[kgts23]

There is a StorageClass-option for veeam k10, to support shallow read-only volumes!

https://docs.kasten.io/latest/install/storage.html#snapshots-as-shallow-read-only-volumes-cephfs-only

@Madhu-1 Thanks for pointing in the right direction!