2012-01-10 09:20 AM
I have been working with support in regards to doing backups/restores of servers with the VSC 2.1.1 plugin. We have been getting the backups to work fine and they are very quick which I like but when doing a restore it has less than satisfactory results in regards to the length of time it takes. A little background, we are on a FAS2020 filer with Vcenter 4.0 update 3 loaded on 2 ESX boxes. The volumes themselves are connected via NFS to the Vcenter.
When I kick off the backups they are very quick and that is not a problem. What happens though is I attempted a Restore of the entire virtual machine and not the individual vmdk files. It took 25 minutes to restore a 25 gb server. Is this normal to take a minute per gigabyte of data? I would think it would be much faster considering the data is simply being restored from one folder to another folder all on the same SAN. It doesn't leave that Filer.
What are others experiencing? I am no expert so I just want to be sure what I should be expecting to be realistic. The reason I ask this question is because currently I am backing up my vm servers with Vranger which is backing them up to an external USB 2.0 drive. I can actually do a restore of the whole server over USB 2.0 and then over the network to the filer in the same amount of time as VSC. That is why I question it.
Thanks in advance.
Solved! SEE THE SOLUTION
2012-01-11 12:45 AM
In an NFS environment VSC uses something called single file snap restore (SFSR) to perform recoveries of objects within a datastore. Unlike volume restore which is more or less instant, SFSR takes time to run. I haven't heard plans to improve it because it will be phased out in favor of using block sharing techniques (dedupe) to achieve a better result. In Data ONTAP 8.1 FlexClone for files, or from the CLI the clone command set, has improved allowing (a) instant clones and (b) clones created from snapshots. So instead of doing a SFSR we can do a file clone. You can do it by hand in 8.1, and I imagine the NetApp mgt products will take advantage of it as well over time. If we look to 8.1 cluster-mode the SFSR engine is gone; if you snap restore a file under the covers it will use the file cloning / deduplication technology.
So far my answer probably doesn't help you that much except to say what you see is within the realm of expectations, and that longer term there are solutions.
For your current situation I would instead suggest a different recovery process that a colleague of mine has written up here:
Hope that helps!
2012-01-11 07:21 AM
This looks to be an extremely viable solution. I really appreciate you replying to this and giving me the extra manual on the restore process. No one has been able to point me in a better direction yet. I will try the actual addition of the hard disk tonight to see how quickly it works and reply back to let you know the results. I did get all the way through the mounting process and did go through some of the steps of adding the hard disk but didn't committ the changes yet. Again, thank you very much.