vmkernel (used for NFS mount) throughput....

[jamieorth]

Ok, I have searched the forums on this but I do not have a good understanding just yet... We are using a NFS mount (DataDomain Appliance) for vRanger backups. We have had issues with the NFS mount dropping so last night we isolated the traffic to a different physical switch. The switch is FastEthernet, not Gigabit. I understood the vmkernel traffic to be throttled at 100M but looking at the traffic from last night we were only averaging 3259 KBps were prior nights on a Gigabit interface (and the vmnic associated with the vSwitch connecting at 1000 Full) traffic was 19756 KBps (spiking). So, is the vmkernel a percentage of the available perhaps??? Needless to say our backup jobs did not finish durring our window - the good news is that the NFS mount never dropped.

Regards...

Jamie

If you found this information useful, please consider awarding points for "Correct" or "Helpful".

Remember, if it's not one thing, it's your mother...

[jamieorth]

No answers?? Someone has to know this? I have found information about the Service Console being capped....NFS throughput from a VM to a NFS mount is much faster than from ESX itself.

Regards...

Jamie

If you found this information useful, please consider awarding points for "Correct" or "Helpful".

Remember, if it's not one thing, it's your mother...

[kjb007]

On your gigabit setup, did you use auto, or did you force your speed/duplex? I've had issues with autoneg, and found forcding to 1000 Full helped clear up some issues.

-KjB

VMware vExpert

[jamieorth]

KjB -

We have had the ports configured both ways at times. When I first got here everything was set to 1000/Full on both the hosts and the switch. IEEE spec indicates not to force Gig so we changed it to Auto on both sides. Didn't make a difference in our case, but we did verify that all pNics and switch ports matched. We ended up changing them back to 1000/Full. Not sure why but that came from Networking so I let them do it....

Regards...

Jamie

If you found this information useful, please consider awarding points for "Correct" or "Helpful".

Remember, if it's not one thing, it's your mother...

[kjb007]

Wow, I think we've had this conversation before.

Can I assume the console memory has been increased, and that you are not seeing memory issues on the service console?

-KjB

VMware vExpert

[jamieorth]

Yeah, I have had several threads open as I try to solve my NFS / vRanger / DataDomain backup problem. Each area I investigate either confirms the issue is with another area or leads me down a darker (and more puzzling) path. Now that I have a couple of months of backups and documented changes along the way I have noticed that the failures only occur on 5 particular VM's (out of 12 that we backup this method). I haven't seen anything that would cause the failure, especially when I backup the same VM through CIFS it works....

The search continues.......

Regards...

Jamie

If you found this information useful, please consider awarding points for "Correct" or "Helpful".

Remember, if it's not one thing, it's your mother...

[kjb007]

What is the size of the vm's that fail? Are you using file-based or image-based backups?

-KjB

VMware vExpert

[jamieorth]

VMDK sizes anywhere from 20G up to 473G. Most of them are 60 - 80G in size with the VM's having multiple VMDK's. Using vRangerPro as the backup software and not using compression on vRanger becuase the DataDomain appliance handles the compression / de-dupe.

Regards...

Jamie

If you found this information useful, please consider awarding points for "Correct" or "Helpful".

Remember, if it's not one thing, it's your mother...