VDR Using 100% CPU and a large amount of memory

[helloitsrainn]

My lab infrastructure has 6 VMs across two hosts plus a host running VDR in a VM (so 3 hosts and 7 VMs total).

I created a basic backup set and chose to back up all the VMs to a machine running Win 2k3 with a basic share. This machine is not virtualized.

While VDR is running and backing up the VMs, the cpu spikes and utilizes ALL of the cpu on the host machine that the VDR VM is running on and I start to get email warnings. Why is this? Can I tame this thing somehow? In addition the basic VDR VM that came from the download via your website includes 2 vCPUs and 2048MB of RAM. Is all of this neccessary? The host machine the VDR VM is running on is a basic Dual Core with 4GB of memory. I didn't think it would need that much. Did I underestimate? Should I just turn the warnings off and dedicate this entire machine to just running the VDR VM?

Thanks

[admin]

Some suggestions

0) Please ensure you are on VDR 1.0.2 - this was posted last week

1) Use a s VMDK vs a CIFS share as your destination. When using CIFS, all the data transfer and dedupe operations needs to run over the network - slower and more resource intensive

2) Look in the logs - how long do your backups take? Not sure that I would dedicate the host just for VDR. I would not change the VM properties - since all of the QA was done with these default VM properties

[helloitsrainn]

Hello! Thanks for the response.

Here is a snippet from my log:

9/15/2009 6:47:26 PM: Starting VMware Data Recovery, version 1.0.2.562

9/16/2009 2:56:20 PM: Normal backup using Backup Job 1

9/16/2009 2:59:21 PM: Copying apps

9/16/2009 2:59:35 PM: Performing incremental back up of disk "[Ravens 300gb|http://communities.vmware.com/community-document-picker.jspa?communityID=&subject=Ravens+300gb] apps/apps-flat.vmdk" using "Network"

9/16/2009 4:49:12 PM: Task completed successfully

9/16/2009 4:49:12 PM: Completed: 5 files, 30.1 GB

9/16/2009 4:49:12 PM: Performance: 280.3 MB/minute

9/16/2009 4:49:12 PM:Duration: 01:52:52 (00:03:14 idle/loading/preparing)

Adding it locally to the VDR VM makes it much harder to get that data

onto tape in a production situation. Any other ideas why it would be

spiking so much CPU?

Thanks,

[toha]

Any other ideas why it would be spiking so much CPU?

It just is like that, deduplication process is very resource intensive.

[helloitsrainn]

Understood. I guess I'll try and adjust the warnings I get from that host.

Finally, is there a way to have VDR email me if there is a problem or whenever a job completes?

Thank you.