VMware Solutions Discussions
VMware Solutions Discussions
I'm new to both NetApp and CommVault worlds, which will become obvious in a moment.
We have a V3140 that we use for backend storage (FC) for our vSphere environment. After playing with SMVI 2.0 we really started to hate VCB more than we did before. However, VCB gives us the ability to backup VMDKs to tape or whatever for archiving and DR. So the newbie question is, before we go and file for divorce from VCB, how do we do a DR backup of our VMDKs with Simpana once a SMVI snapshot is taken? Is this even possible or do I need SnapVault or something else in place?
how do we do a DR backup of our VMDKs with Simpana once a SMVI snapshot is taken?
CommVault supports NDMP to tape & SnapMirror to tape (SM2T) - in both cases you stream a particular snapshot directly to a backup device.
The real trick of course is how to restore a LUN backed up through NDMP, I am still looking for a nice description on how to do this effortlessly... You will have a NetApp LUN 'bubble' around your VMDK files, so as far as I can tell there would be no easy way to restore a single VMDK from tape.
The real trick of course is how to restore a LUN backed up through NDMP, I am still looking for a nice description on how to do this effortlessly...
It may be a tad tricky, but definitely doable - have a look at this thread re NetBackup:
I should have mentioned that our filer is just presenting SAN LUNs to vSphere. We're not using it for NAS or NFS or anything like that. I'm not sure if that makes a difference with NDMP or not.
As mentioned, having LUNs instead of an NFS volume is different. This has to do with the level at which you make the backup, with NFS you are directly in the file system backing up 'VMDK' files, as opposed to LUNs - which will look like 1 big file from which you cannot directly restore a VMDK file without recovery the entire LUN first. (at least when it is on tape). Please test this before you implement, try to contact your CommVault rep to give you some pointers on the restore process.
We were in exactly the same situation, SMVI, CommVault 8, etc.
We moved away from FC to NFS on dedicated switches. So much easier to manage, cheaper, no HBA, fiber cables, fiber switches and you can restore everything you need from NDMP backup because it is open file system, you can just restore one VMDK or all, whatever you want.
Performance is the same, no downside. We have 2 dedicated HP Procurve 2910al (10gb up, 1gb down) uplinked to FAS2050 via 4 10Gb cables, and each ESX host has 2 1Gb connections to switches.
If you can, move away from FC to NFS, will save you money and time in a future.
This should also apply to Veritas tape backups right?
Yes, this has nothing to do with the type of a backup software, it is a protocol thing, FC vs NFS
Thanks this really helps.
Some excellent advice in the answers 🙂
Just an observation, particularly since Eric mentioned being fairly new to NetApp: please don't lose sight of the simple fact that you really should make extensive use of the snapshots on the primary disk.
In other words, you use SMVI to take those backups, and recognise how good they are.....so my advice here is to
It will always be quicker to recover from those on-disk snapshots, even if you continue with using LUN's, than performing any tape-based recovery.....and I always remind people the reason they backup is in order to be able to recover!
The efficiency of snapshots (& hopefully you are taking a good look at deduplication with this data - VMware is one of the areas our custies do seem to get the biggest space savings with dedupe!) mean you can probably keep a months worth of backups for relatively little extra disk - perhaps only 10-20% disk......enabling you to have maybe 31 "full" daily backup copies and perhaps even more more recent hourly ones 🙂
If you have the luxury of a replicated copy offsite (think volume SnapMirror), then you will then also have all those copies in 2 places, which will lower your reliance on tape even further.....
I'm a CommVault employee, but I was a PS guy for NetApp for 5 years. I followed this link from Twitter. There are a couple suggestios I would make for you to consider.
VCB is a pain. For immediate relief, see if you can move to VADP. Simpana 8SP4 and later support VADP if you are on vSphere 4.x and have virtual machines with hardware v7. VMware made significant improvements in data protection with VADP to replace the headaches common in VCB. You will be much happier with it.
You can ndmp the LUNs to tape, but this is a bit of a manual process. You would need to recover the entire LUN (datastore) from tape if you ever needed to recover a VM (or even a file within a VM), and that too is a manual process. If you only expect a restore from tape once in a blue moon and just need a "good enough" copy there, it might do the trick. Otherwise, tt might be better to just use VADP and Simpana for tape copies to make recoveries easier and still allow you to use SMVI.
Since you are already a CommVault customer, you may consider using SnapProtect for VMware in Simpana v9 to coordinate the NetApp snapshots and also allow a proxy host to move the data in those snapshots to secondary disk storage or tape. CommVault would allow the seamless recovery from anywhere, without manual steps, and still allow for Full VM or catalogued and searchable single file restores using the CV GUI you already use for the rest of your environment.
CommVault and NetApp make a great pair. It is always nice to see customers who are fortunate enough to have both.
We are NetApp and Commvault customer too. Here is my setup;
All VMs are on NFS Datastores, i use SMVI to do nightly snapshot and then i have a NDMP policy to backup snapshot_recet (most recent only) directory to tape later the same night.
It works pretty well, i can restore entire VM or just VMDK and mount it to another existing VM for single file restore.
However, the only problem i have it that im using thin provision disks, but when NDMP bacs up smapshots all the VMDK file are their "real" thick size, so my NDMP backups are really big.
What would be the better solution for me? i read about the SnapProtect for VM but that requires a proxy, i dont really like proxies, how will iData Agent for VM work with VADP? will i need a proxy? do i have to install an agent on every VM or does it work underneath VM level.
BTW we are on Simpana 9 but vSphere 4.0 (will upgrade to 4.1 later this year)
A proxy server isn't required, even for SnapProtect. Simpana can use your existing ESX servers as the proxy along with a VM as the VSA agent. I know a lot of the example architectures show a separate proxy, but it is not required. The VSA agent interacts with VMware's VADP API directly and backups occur either over the SAN (for enviroments witha proxy) or, in a non-proxy envorinment, the Guest VM acting as the VSA will mount each of the VMDK disks to be backed up directly from the vmfs datastores using "hot-add" mode.
I happen to think it is more efficient as you can use changed-block tracking to only send the changed blocks of the vmdks for incremental backups instead of the full VM each night.
We have Netapp FAS3020 filers in our main site and a single FAS3020 in our DR site. We use Snapmanager for Exchange, Snapmanager for SQL, Snapmanager for Virtual Infrastructure, Snapmirror, and Snapdrive to backup Exchange, SQL, and VMs.
We also use Commvault 8 to backup a few stand alone servers and do an NDMP dump from the filer in the DR site to an Exagrid box for longer term retention. The problem is Exchange and SQL get vmdks dumped to the Exagrid so these backups are really the entire database.
We are thinking about moving all backup to Commvault (Simpana) 9 and not using the Netapp products (other than Snapmirror) for all of our backups. As someone who worked at Netapp and now works for Commvault can you discuss the pros and cons of making this move?