We have two FAS3020s in a cluster and use an NFS datastore for three ESX 3.5 servers that run about 22 VMs. We are just about to deploy SMVI v2 and I had three questions:
1. Does SMVI need VCB?
Right now we use backup software that leverages VCB to backup our VMs. After doing some reading it appears that SMVI does not use VCB. It quiesces the VM using VMWare tools, takes a VM snapshot, and finally a Netapp Snapshot. Once the Netapp snapshot has been taken, the VMWare snapshot is deleted.
2. Can I use SMVI to backup SQL and Exchange?
I have both SQL 2005 and Exchange 2007 running as VMs. I present my Exchange server with two luns from my filer via the iSCSI initiator and use Snapmanager for Exchange to backup my Exchange server database and logs. I present my SQL server with four luns from my filer via the iSCSI initiator and use Snapmanager for SQL for backup. What I am really asking is can I use SMVI to backup the C: drive which is what is encapsulated inside the VMDK? I know SMVI will not backup the luns attached to these two VMs but I do not need it for that purpose, This articled talks a bit about this but I am still unsure:
1)SMVI does not need VCB. As you correctly point out It quiesces the VM using VMWare tools, takes a VM snapshot, and finally a Netapp Snapshot. Once the Netapp snapshot has been taken, the VMWare snapshot is deleted.
2)Since SMVI (with VMware snapshots turned on) relies on VMware quiescing of virtual machines when making backups, it is able to provide – “for free” - application-consistent backup through VMware's VSS requester/provider components for the applications running inside the virtual machines.In fact, all SMVI backups with VMware snapshots turned on are “application-consistent”.
Since the VMware VSS components have no application recovery capability, and no VSS writer-assisted recovery is possible using VMware’s current technology, the only recovery mode available to the application is recovery to the point of the last backup. It is not possible to roll forward the logs, recover to a specific point in time or specific transaction, or have other enhanced recovery functionality. SnapDrive and application-specific SnapManagers (such as SME/SMSQL) running in the guest provide application-consistent backup and fine-grained recovery for applications whose data is stored using Microsoft iSCSI Software Initiator LUNs or RDMs.
Thanks for the quick reply and I have a follow up question. Is my description below of backing up Exchange using SMVI and database/logs using Snapmanager for Exchange accurate?
BACKUP EXCHANGE USING SMVI
I plan on backing up my Exchange server using SMVI which will back up the C: drive but will not back up the two luns I present to my Exchange server which house the Exchange database and logs. For those, I already run Snapmanager for Exchange. Let's say my Exchange server crashes as a result of a virus and I need to restore it. I can restore Exchange using SMVI which would get the C: drive which houses the Exchange application back to the point when SMVI ran last. I would certainly lose items since the last time that SMVI ran but in the case of Exchange that is items like message tracking logs which are not hugely important. After the recovery is complete, the Exchange server would still be presented the two luns from my filer and everything would work fine.
BACKUP USING SNAPMANAGER FOR EXCHANGE
Let's say my Exchange database somehow gets corrupted. I can use Snapmanager for Exchange to recover the database and logs but I will lose any transactions (email, calendar items) since the last time I ran SME. Once the recover is complete, the database and logs are presented to my Exchange server via two luns and I am back in business.
3. Does SMVI have any problem when used against domain controllers?
You can use SMVI for back ups, but be very careful with restores.
DCs have continuous conversation between them & conflict resolution mechanism. So say, you've done something nasty to your AD (e.g. deleted all users) & then you simply restore from snapshot your DC1. DC2 will then talk to DC1 and say: "hey, my records show more recent changes to yours, so please follow what I have on my books". So after a while you are back to square 1, i.e. the state which was before rolling back DC1.
To avoid this you need to do authoritative restore on any of DCs. You can still use SMVI-generated snapshots, but you simply mount one of them to allow basically a simple copy of AD database via NTDSUTIL.
With either method you have the same problem Radek mentioned which is an out of date domain controller. I would default to just suing SMVI then depending on the restore needs decide how I want to restore. (it might just be easier and safer to deploy a new DC than recover one from some time ago). I keep pushing the idea of using SMVI to take hourly backups without VMware snapshots. The VMs are obviously not is as good of a state but you then have a much shorter recovery point. With a DC this could be a very good thing as most DC will resync as long as they are not more than an hour out. If you only keep 2 hourly backups the cost of these backups is very little disk. In my lab 3 hourly of about 15 VMs uses about 700MB of disk.