Hi folks -- I'm tired of some users constantly using 'snapshots' as a backup mechanism, and it's tedious to manually clone or backup vmdk's for them, so I thought I'd give VMWare's "Virtual Data Recovery" tool a try, in my ESXi 4.1 Enterprise environment.
Seemed straightforward to deploy and configure (a bit of Googling needed to find the initial UID/PWD), and as suggested by the Docs added a 2nd 'hard disk' to the appliance to serve as the storage repository: a 1.5 TB disk created on an NFS-shared NAS volume.
I selected about 20 vms, chose 'today' in the schedule so it would kick off, and -- boom. System hangs after 15 minutes with high-cpu/memory warnings. I finally get the guest OS to shut down (take a while to clean-up all these inflight operations) and I power it off and reconfigure it with 2x the vcpu and RAM: 4 and 8. But on powering up, again I get warnings and the system in unresponsive. I don't see what I'm doing wrong. The VM's/VMDK's I'm backing up aren't tiny at ~50GB each, but....
It is normal for a VDR (and also for VDP) appliance that you need all the resources you will give the VM for the very first backup. Dedublication is a very CPU intensiv task especially for the fist full backups. I was using VDR with 4 CPU's and 32 GB RAM for the fist backup, than i reduced the CPU count to 2. More memory will always be better. If you have troubles with such a high load in your environment, you can reduce the IO-Load by defining an IO-Limit for the appliance. You can also reduce the number of backups which will run in parallel by setting the parameter "MaxBackupRestoreTasks" in the datarecovery.ini file. (see here: VMware KB: VMware Data Recovery: Enabling verbose logging and datarecovery.ini options ).
I was using a limit of 500 IOPS for the backup-destination disk to reduce the SAN-Load of the backups. The warnings in vCenter will be normal for a VM with such a high CPU-Load, the unresponsiveness may be a fact of the overloaded destination device.
After the fist full backups the load is getting lower, but you will always have a high CPU-Load with dedublication. I would first try to set the parameter MaxBackupRestoreTasks to a value of 2 and if this doesn't help you can also try to reduce the IOPS for the VM with "Edit Settings" - "Resources" - "Disk" - "Limit - IOPS".
Thanks Paul -- all in all, everything seemed to run and complete. I'll play around with the settings.