Performance issues on virtual machines stored on NFS datastore

[tis]

Hello,

Our configuration :

NetApp 2050A

ESXi 4.0.0,208167 with Virtual Lab Manager

Swich Gigabit (cisco catOS)

Static Multimode IP load balancing between filer and switch

No multimode on the ESX.

2 NFS datastores stored on SATA aggregates

1 NFS datastores stored on FC aggregates

We have applied all recommendations described in NetApp technical document TR3428.

Our users experienced very low responses using virtual machines hosted on filer NFS datastore. Quite often the vitual machines even freezes for several seconds.

The behaviour is the same on both SATA and FC aggregates.

For investigation purpose we have built a linux-based NFS server. User experience with the virtual machines stored on the linux server is quite good. There no performance issue here.

When the virtual machines stored on the filer are freezing, virtual machines stored on linux server operate normally.

The filer does not show any CPU, Disks or Network I/O saturations. The ESXi server does not show any resources saturations either.

Is there a way to monitor the throughput (Bytes/sec) used by a given virtual machines? or the throughput only between the ESXi and the Filer?

Could someone help, please?

Thank you.

Ruben.

[kevingolding]

Hi Ruben, do you have any SMVI jobs running at the same time?  and have you configured the NFS/TCP changes in the advanced configuration mentioned in the vshpere on Netapp TR Doc, especially the TCP/IP heap settings?

The other thing to check, is the disk alignment, if the VMDK's are out of alignment the more and more VM's in a datastore the more the problem is prominent...

Other than that, check the network settings at the switch, flow control etc can cause a hassle...