Is it possible to create a VxVM Cascaded Space-Optimized snapshot from a Netbackup policy?

[Symen]

Hi,

I'm trying to figure out if it is possible to create a VxVM Cascaded Space-Optimized snapshot from a Netbackup policy. The Snapshot client documentation indicates that it's possible to trigger a FS Checkpoint, and also Space optimized snapshots, but what about cascaded snapshots?

[RLeon · Accepted solution]

For VxVM based snapshot methods, regardless of whether it is for NetBackup's Instant Recovery feature, or for the more traditional backup-from-the-snapshot purpose, the snapshots that NetBackup instructs VxVM to create are always "independent" snapshots. That is, the snapshots are not dependent on their neightbouring siblings, but only to the orignal volume. (Or even not, if using full-sized instead of space-optimized snapshots.)

The snapshots created by VxVM's Cascaded Snapshot feature, as well as the Snapshot of a Snapshot feature, are not independent snapshots; they depend on each other. And as documented in the Nbu snapshot guide, NetBackup does not create this type of snapshots through VxVM.

Side note:
One would argue that VxVM's Cascaded Snapshots - most likely redirect-on-write instead of copy-on-write based - is also a type of "Snapshot of a Snapshot" feature. But, alas, "Snapshot of a Snapshot" is already used to refer to something else specific.

[Symen]

Thanks for the reply! That answers the question. It does bring up a new one though:

If I created 12 simultaneous snapshots (one every two hours), does that mean each write operation to the volume will result in 13 writes to the backend disk?

[RLeon]

That is correct. The 1st being the original volume, and the rest of the 12 being independent snapshot storages, each would need to be updated... independently. In other words, the more independent snapshots that exist concurrently, the slower your writes. That is exactly the kind of scenarios that Cascaded Snapshots - with its faster writes - could help.
The situation gets kind of reversed if you use full-sized snapshots, where each break-off means one less snapshot to feed.

Either way, you may want to go easy on the Maximum Snapshots parameter, by simply backing up some of them instead of leaving them on storage.