Multiple Nimble Microsoft VSS snapshots?

[julez66]

I've not tried this yet, I believe I know the answer, but lets say you have 3 or 4 volumes for exchange or sql databases.

Normally I'd say create a volume collection and have them be under 1 protection policy.

Since a verification over a 1Gb link can sometimes take awhile.

Could you split the 3-4 volumes into their own protection policies and run them at the same time?

Or is the Nimble VSS writer going to be tied up with a single job at a time?

The volume collection of course seems to do 1 volume in a collection, finish it, then move on, and so on.

[#2]

Hi Julez,

Nimble does not use our own VSS writer. We use a requester through our Nimble Protection Manager (NPM) that simply leverages the native Windows VSS writer. The "single job at a time" is a limitation of the MS VSS writer. So, you could split the volumes into their own volume collections and schedule them all to occur at the same time, but Windows would still queue of the requests and do one at a time. All this would buy you is more work with no apparent benefit.

You could stagger the snapshots, but then you would have to worry about data consistency. You may want to consider putting each database/log volume pair in a Volume Collection of their own. That way you are staggering the workload somewhat while maintaining consistency between the database & logs.

Hope this helps.

Brandon Grieve

[bgrieve65 · Accepted solution]

Hi Julez,

Nimble does not use our own VSS writer. We use a requester through our Nimble Protection Manager (NPM) that simply leverages the native Windows VSS writer. The "single job at a time" is a limitation of the MS VSS writer. So, you could split the volumes into their own volume collections and schedule them all to occur at the same time, but Windows would still queue of the requests and do one at a time. All this would buy you is more work with no apparent benefit.

You could stagger the snapshots, but then you would have to worry about data consistency. You may want to consider putting each database/log volume pair in a Volume Collection of their own. That way you are staggering the workload somewhat while maintaining consistency between the database & logs.

Hope this helps.

Brandon Grieve

[julez66]

I would think that 4 active databases and 4 archive databases with a total of 1000 users would be pretty small potatoes, even for the 1GbE arrays.
However I think due to MPIO being unusable during verifies it REALLY hurts that replay performance.
With the help of Microsoft and VMware we architected the system to have 4 separate volumes (active, archive, active logs, archive logs) rather than 16 so splitting those up with different protection policies really isn't an option at this point.  Also not sure this would solve our issue due to the VSS writer limitation anyway, because it's essentially staggering the snaps for us anyway.

We didn't think that the verify of the 4 archive and 4 active databases would take 8+ hours to complete...