I was reading an article today (http://www.integrateddatastorage.com/blog/2012/01/10/snapshot-or-not-to-snapshot-that-is-the-question-when-leveraging-vnx-unified-file-systems/) on the IDS blog, asking whether to snapshot a filesystem stored on a VNX array. The quick answer is depends. Yes, create snapshots and use that as one of the many tools you have at your disposal for recovering a users data back to a point in time. Yes, create point in time copies of data on independent hardware (a backup) so that you can recover that same data no matter what happens to the source. Yes, replicate those independent copies to a second location so that you can survive a local catastrophe with your data intact, and yes, snap and then replicate your data on your VNX to the same or different remote location. Just don’t depend upon a snapshot to get you out of every jam. The very first limitation listed in the “Using Snapsure” manual is
“A checkpoint is not intended to be a mirror, disaster recovery, or high-availability tool.
It is partially derived from realtime PFS data. A checkpoint might become inaccessible
or unreadable if the associated PFS is inaccessible. Only a PFS and its checkpoints saved
to a tape or an alternate storage location can be used for disaster recovery.”
That puts it fairly succulently I would say, but let’s ask another question around retention. Since a snapshot is so good (and fast) at creating a PIT copy, this is a tool I would want to use throughout my business day. Conservatively, let’s say I take a snapshot to start the day, one late in the morning, another around lunch, one in the early afternoon, and then one at the close of the day. That’s 5 every 24 hours. Snapsure supports 96 read-only snaps, so that equates to about 19 of the 24 hour periods that I can retain data. How do I get data back that is 60 days old from my snapshots? I can’t…that’s where our good friend Mr. Backup can assist. What about storage costs? I know that shiny new VNX was very affordable, but that is your primary storage device. Your entire datacenter runs off of it. If you are holding 96 snaps, reflecting 19 days worth of data, then you are also holding 19 days worth of extra information and consuming precious storage in the meantime. Let a purpose built backup appliance help with some of that burden on your storage. That’s what it was purpose built for after all. What about corruption? Us backup guys always talk about corruption, but it is something you have to think about. Corruption can be accidental, but it can also be malicious, and in most cases, that snapshot is going to just keep on snapping regardless of the data’s integrity. Mr. backup can get you back to a state prior to that corruption. What about mobility? Where can that snapshot be restored? Back to where it was taken from. Where can my backup be restored? Anywhere I please so long as my backup software can access it. That gives me some real flexibility.
In short, Snapshots are awesome and I encourage everyone to make them part of their data protection strategy, but they are not the panacea for all that ails us. Thing holistically about your data, because it is likely the most important thing in your datacenter.