Snapshot policy Best Practice

[AlexeyF]

Hello

Is there some guidance on a reasonable number of snapshots that could be kept for a volume without considerably impacting performance?

As an example: if I configure snapshots policy to keep 15 weekly, 30 daily and 100 hourly snapshots, will it create a lot of extra load to this volume and increase the response time?

Thanks

[paul_stejskal · Accepted solution]

It really depends on your system (model and # of drives if HDDs), what the frontend workload is, etc. This might be a question for your account team. One thing you could do is SnapVault to another cluster and retain snaps there.

[EWILTS_SAS]

The challenge with snapvaulting to another cluster is that restoration is harder - it typically requires a storage administrator.  If only NetApp could provide a seamless way for end users trying to restore their own data to walk the snapshots to the other cluster and pull the data from there (assuming the export policy allows it).  After all, you KNOW the relationship exists but the users may not (almost certainly won't know the cluster name or svm name).

[EWILTS_SAS]

Just imagine the power of doing a cd .snapshot, finding ALL of the local and remote snapshots, and doing your own restore.  We're already doing seamless recovery from data tiered to an object store and this would take it a significant step farther.  Yeah, it's hard - that's why we pay you the big bucks 🙂

[paul_stejskal]

I understand. If you don't have a high amount of workload this could work on the primary, but you probably need to talk to the account team.

Once an hour isn't too bad.