2009-03-11 05:29 AM
I am new to using NFS. I have created four NFS volumes presented to our FAS3020C filers. The NFS volumes are roughly 200GB each. I have a few development vns running inside of each NFS volume. I currently have snapshoting configured for 14 days. What is 'best practice ' to restore a single vm? If I perform a 'revert to snapshot' then all vms in the NFS volume will get reverted. Can I simply copy out the snapshot I want to revert to? Will I have to delete the present vm?
Thanks in advance!
Solved! SEE THE SOLUTION
2009-03-11 06:17 AM
If you are not using SMVI, then you can just use the virtual infrastructure client to browse the datastore, delete the vm, then copy/paste the vm from the snapshot version that you want to revert to back to the active filesystem. Hope this helps,
2009-03-11 06:48 AM
Thanks for the response.
Is that the 'quickest' way? I experimented with a vm with a 10GB vmdk and it took 10 minutes. Obviously it would take considerably longer, given the size of the vmdk.
2009-03-11 06:57 AM
Unfortunately, the answer to this is "it depends". It can be faster to use single file snaprestore (snap restore -t file) on all files in the VM directory, but that can also be slower in some situations.
In particular, if you are using deduplication, a single file snaprestore can sometimes take a bit longer than a copy out operation if you are running 188.8.131.52 or 7.3.1. This should not be the case though, and single file snaprestore performance should improve when 7.3.2 is released later in the summer.
BTW, SMVI and VIBE both automate the process of running single file snaprestore on all the files in the VM directory, so check out those tools if you want to go that direction, as it will make your life much easier.
2009-03-11 07:20 AM
As Mike said SMVI automates this restore but also lets you do a neat trick which allows very fast restores. In the SMVI interface you have the option to "mount" the backup. This makes a writeable version of the backup and mounts it to a ESX host. Takes about 5-10 secs. Once that is done you can browse to the new datastore, register the VM you want to restore and power it up. The VM at this point is running on the backup but is back online and users can begin to use it again. Essentially it is "restored". Finally you can use VMware's Storage VMotion to move the VM off the backup onto the production NF volume. This will take some time but happens behind the scenes. Once the storage VMotion completes, you can unmount the backup. The restore isn't actually faster but to the end user it appears like a very fast restore
2009-03-11 07:22 AM
If you use snap restore -t file (along with the other required arguments) then only the file given will be restored, and all other files in the volumes (including all other VMs) will be unaffected. Using snap restore -t vol would revert the entire volume, and all VMs in the volume, back to the snapshot point in time.