ESX 3.5 on NETAPP NFS Performance problem

[miklaw2]

We are running ESX 3.5 Build 158874 on 14 hosts running total of 56 Vms with the vms being on 3 different NFS datastors

They are divided something like this

FC1 contains 8 Vms

Sata1 contains 34 Vms

Satatmp contains 14 vms

Filer is a FAS 3050c now running Ontap 7.2.6.1.

I have been implementing the TR-3428 manual after the fact... and there is still some stuff undone such as the partition alignment for the windows servers

We were running fine until the end of May...when we started seeing a severe slow down on our VMware servers and overall everything on the filers...Virtual Center says all is well.. but the servers are moving in slow motion. We were on OnTap 7.2.2 and we upgraded to 7.2.6.1 this past Saturday but we are still seeing a performance hit...though nothing is showing up Filer cpu is fine etc... the servers on the SATA1 volume is slower then the others... but 34 vms should be ok according to the documentation

We note NFS traffic has gone up (but we expected that) so any thoughts/ideas/ something I can check... Filer side or VMWare side....

Thanks for anything you can offer.

Mike

[RParker]

I have been implementing the TR-3428 manual after the fact... and there is still some stuff undone such as the partition alignment for the windows servers

You didn't state type of drives, I assume from the names you gave the NFS datastores they are SATA. If so, that's where your bottle neck is. SAS drives are much better performance...

but 34 vms should be ok according to the documentation

For certain tasks and where do you see this documentation? SAS drives, the number of spindles, the type of RAID all have an impact.. What are the types, size and configuration for the drives?

[miklaw2]

I asked the same question in the NetApp community and they believe it might be related to a aggregate being over 90% ...which did happen recently ... we are going to work to fix that and see if that fixes the performance problems.

To answer your questions though we have a mix of SATA750s drives and FC300s we are seeing performance hits across both...just more greatly on the SATAs

The documentation is from NetApp TR-3428 Netapp and VMWare Virtual Infrastructure 3 Storage Best Practices.

I think this is solved with the aggregate problem.. we will address from that direction.

Thanks!

Mike

[dxb]

Try running "sysstat -u 5" on the filer and watching the "CP time" column. We've had problems where pending writes to slower SATA disks were causing bottlenecks in the filer, which manifested as a string of 100% values in that column.

[miklaw2]

Thanks... we will give that a try.. .

Mike