Unsolved
This post is more than 5 years old
39 Posts
0
604
Advice Needed
We are running 6.1.0-402 with 10 3.9TB nodes along with vcenter intergration. I have a large "File" server (VM) that is about 5TB in size. I chose to backup the system from the Windows level using the agent due to its size. This is going fairly well after some tweeking but does take a long time to backup nightly. This got me thinking about restoration times in the event of a disaster. This server would take forever to restore!
Would it be a good idea to split the server up into smaller servers? If we do this, is it better to perform image level backups? Are restores faster with FLR?
Do I have any other choices?
Thank you in advance.
ae86levin
70 Posts
1
June 2nd, 2013 20:00
We have multiple large file servers in our environment. I back them up as client rather than VM snapshot mainly because we have raw device mapping on some of them (which can't be snapshot). My experience with trying to do snapshot backups of large file servers (the ones without RDM's) wasn't good. I don't recall the specific problems as it was over a year ago but I think they were slower. In any case I decided to go back to client backup.
All of our non file server VM's are backed up as snapshot.
Rather that splitting into separate servers you can backup in separate datasets to split the backup up. I find this helpful for a number of reasons:
1. It means that if there is a problem completing the backup within the backup window you will get some complete backups where if it is a single large backup that times out you get nothing in the way of a recoverable backup.
2. It allowed me to specify separate cached files for the backups which can improve performance and also will allow you to run multiple concurrent backups from command line if you ever need to.
3. It made replication to our off site DC easier as multiple smaller backups are easier to get replicated in 7 days than single large ones. The partial replication data 'falls off' after 7 days meaning if you can't get the whole backup replicated in 7 days you start again.
4. If you have shares that don't change much and will complete quickly (eg archive) then these can be run earlier and guarantee that they will complete rather than being stuck behind shares that take longer.
I have found FLR restores from VM snapshots to be slower to perform - certainly the action of browsing to the file that is to be restored is slower in my experience.