Unknown NFS ops on NetApp filer

[BOERLOWIE]

Hi,

we have a NetApp Filer and i see a lot of IOPS on both controllers.  Using Management Console, i was already able to narrow it down to specific volumes which are NFS Exports.

When i look at the performance statistics, it shows about 15k IOPS on average constantly.  Using Management Console, it shows as all of them as 'nfs_other_ops'.  I'm wondering where they come from and if i can narrow it down further...

We run OnTAP 8.0.2 (plan to upgrade to 8.1.1 in the next few weeks).

thanks a lot!

[thomas_glodde]

hey,

you could enable nfs per client stats (options nfs.per_client_stats.enable on) to see where those iops come from (nfsstat -h or nfsstat -l).

On a personal note, go for ONTAP 8.0.5 instead of 8.1.2 ...

Kind regards

Thomas

[casejohnh]

Check out the following link - https://communities.netapp.com/thread/6011

[virtrixke]

Thanks,

i found the hosts with this command.  It were some ESXi hosts that were doing massive amounts of 'Lookup' operations (even though nothing is being done over NFS from that ESXi host).  I have opened a case with VMware.

The other hosts are RHEL which are doing 'getattr' and 'access', again by a massive amount.  Our Linux guys are investigating those hosts.

Thanks!

[virtrixke]

Thanks,

opened a case with VMware as the NFS other ops are coming from ESXi for an unknown reason.

[igalkatzir]

just a quick note,

When I could not run the 'nfsstat -l' since it has a vfiler enabled.

hence, I needed to run the below command instead :

vfiler run vfiler0 nfsstat -l, which resulted in -

===== vfiler0

10.10.97.157    <hostname unknown>              NFSOPS = 30013399682 (100%)

10.10.97.217    <hostname unknown>              NFSOPS = 27522169102 (100%)

...

does anyone knows if this is an aggregated value of all Ops done by this client since the stats where zeroed?

Igal

[bbjholcomb]

Could this be nfs other ops? We see a lot of these, most don't go to disk, NFSV3 does a lot setaddr and getaddr.

[JoeF]

This is an older thread, but can you provide an update on what VMware was able to tell you? 

I'm seeing similar behavior on our vmware NFS exports.

Thanks,

Joe