Video Screencast Help
Symantec to Separate Into Two Focused, Industry-Leading Technology Companies. Learn more.

Question on Resource Limit with VMWare snaps.

Created: 16 Jul 2012 | 2 comments

Fellow Backup Admins,

At our company we seem to be running into some sort of resource contention with VIP's.  My question is if anyone knows/has references to what the snap limitations are on the datastore level, cluster level, and datacenter level.  There is no error to provide at this point, only symptoms which is vm's having multiple snaps of the same snap.  I did try to search the forum, but the best match was from 2009.

This would be to help tune the global settings for VMWare resources. (See http://www.symantec.com/docs/HOWTO70967 for more information.)

Thank you all for the assist.

Environment Info:

NBU 7.1.0.3, RHEL 5 Master

NBU 7.1.0.3, Win2k8 Media

vCenter 4.5.x

Comments 2 CommentsJump to latest comment

cbode's picture

unfortunately there is no hard and fast rule since the snap creation process is so hardware dependent.

Things to consider

1) Are you using hardware assisted snapshots?

2) are attempting to snapshot multiple VMs residing within the same datastore at the same time?

3) how beefy is your vcenter server?

GnuubBackupAdmin's picture

1) Are you using hardware assisted snapshots?

A) I'm actually not sure what you mean here?  It is via SAN option not NBD... But no funky array side snaps.

2) are attempting to snapshot multiple VMs residing within the same datastore at the same time?

B) Yes.  We have clusters that rise up to 500+ vm's.  A Serial approach isn't going to help us much.

3) how beefy is your vcenter server?

C) vCenter itself isn't too beefy fairly older box.  I don't know the exact model and CPU/Memory layout, but are you referring to the ESX hosts themselves?  Those range from Dell R810 to R900 generally maxed CPU/Memory.

And a number would be nice, but a formula would work fine as well.  I mean if there was a guideline that X hardware can yield Y snaps then at lease some extrapolation could happen... But there doesn't seem to be a straight guideline to answer this question at either SYMC or VMWare.  At least that I have been able to dig for.

BTW, thank you for the input, cbode.