High CPU Loads With WHM/Cpanel Backups

[Evcthosting]

We are having some issues with WHM backup taking long periods of time at high CPU usages.

We are currently using ESXi Vmware running Guest Machines on dell poweredge 2850's. several flavors of Dual/Quad Xeon Cpu's (3.2Ghz , 3.4Ghz, 3.6Ghz) most have substantial amounts of compatible ram and have the proper resources allocated for each machine.

here is what we see happening: just one example

(6999 username 39 19 4564 1092 908 R 7.2 0.0 2:47.13 gtar) = almost 3 hours to Tar up a 512mb site/account

we have set tweak setting / cpus accordingly as well as running the backups at 1 am Est. with low load times. We have changed the setting to do inc and non inc as well. Our version is cPanel 11.25.0-S45750 - WHM 11.25.0 - X 3.9 (STABLE tree).

Tried Ftp & local backups ect and get the same results on several different hardware platforms. We have tried several different WHM releases with little result.

If anyone in the community has some tips or ideas please let us know...we are open to try them out, aside from rsynce, R1 and other third party programs suggestions, we already have most of that in-place but for our Dedcated/VPS users we want to be able let them do this them selves using the built in script that cpanel has provided.

thanks for any help in advance!!

 

[garrettp]

I would suspect your CPU/memory is more than sufficient, however backups are more of a disk I/O intensive processed and since you didn't mention anything about your disk setup, I am going to assume that is the bottleneck. During the backup process, run top and check the %wa column under CPU. This tells you the amount of time the CPU spends waiting on disk I/O to perform operations. If it's more than 0-1% under normal load or >10% under backup load, you probably have an issue with your disks being your limiting factor.

I'm not too familiar with VMWare, however it may have an I/O scheduler that you would need to play with that could be causing the node to artificially see a disk I/O bottleneck.

 

[Evcthosting]

Thanks

i suspected the disk I/O to be intensive on this process and i think you may be correct. we are running Raid 5, 6 15k Ultra 320's on most of our box's.

as far as the Esxi (vmware) host is concerned, i will look further into the advanced setting on that. "bottlenecks" happen using a "managed" system like this and if there are I/O limits it will become apparent.

I know that 11.25 (version of whm) has had some load issues while doing backups and was not sure if this one of those things were we are waiting for a new roll out on the WHM (version tree) or something else. I will investigate on our off peak hours this weekend and report back....i may still need advice for sure.

i have attached a graph of our test vps IO loads as well.

thanks garrettp!!

http://evcthosting.com/images/io/screen.png