[clue] troubleshooting nfs mounts?

Chris Fedde chris at fedde.us
Wed May 8 16:29:38 MDT 2013


Glad you found a solution.  The other item I forgot to mention is that
rsize and wsize on nfs mounts are a big part of optimizing performance.
 "rsize=8192,wsize=8192" in the mount options in /etc/fstab or the
autofs/amd configs depending.


On Wed, May 8, 2013 at 3:37 PM, Mike Bean <beandaemon at gmail.com> wrote:

> I'm sorry I forgot to respond to this to say thank you for the advice and
> suggestions. Ultimately what ended up happening was the server owners
> agreed to let us end run around the problem.  (They got a mount on a
> significantly higher performance storage device, thank god.)   I honestly
> wasn't looking forward to perf tuning a  1GB pipe so much it squeaked only
> to be told it was too slow.
>
> Mike Bean
>
>
> On Fri, May 3, 2013 at 6:22 PM, Chris Fedde <chris at fedde.us> wrote:
>
>> First all the basics.  Are any of the interfaces counting errors?  Are
>> the interfaces properly configured to auto configure full duplex.  After
>> that start using the dd if=/dev/zero of=/path/on/nas and dd if=/path/on/nas
>> of=/dev/null to get a basic idea of the streaming performance.  Compare the
>> rates you see this way with the theoretical max.  Do similar things using
>> nc to characterize the network performance
>>
>> The optimal throughput you can expect on a 1Gbit network interface is
>> about 120Mbyte/sec.
>>
>> After you have all that working and tuned properly get nagios or another
>> NMS running and start monitoring to get a feeling for what is normal and
>> what is exceptional.
>>
>>
>>
>> On Fri, May 3, 2013 at 1:40 PM, Mike Bean <beandaemon at gmail.com> wrote:
>>
>>> Peculiar problem, haven't the slightest idea where to start.  Need
>>> advice!
>>>
>>> RHEL5.4 box with an NFS mount point to a NAS storage device into a NAS
>>> device with jumbo frames disabled.  NAS has a samba share capability.
>>>
>>> Users are convinced it's too slow.
>>> At least one user is convinced that if he uses smbclient to untar the
>>> same tarball, it takes, 2 maybe 3 minutes.
>>> (My manual untar attempt of the same tarball has been running for 25
>>> minutes and climbing.)
>>>
>>> NAS is in a one-volume raid 6 - seven drives.  Users not showing much
>>> interest in splitting it into multiple volumes.
>>>
>>> Network guys are convinced the slowness is because it's behind a 1GB
>>> router;
>>> Not seeing anything that really worries me in /var/log/messages
>>> Not seeing any unusual processes in top
>>> memory use isn't excessive - not swapping
>>>
>>> network mtu is at 1500 - wouldn't matter, since our switch ports are at
>>> 1500 anyway,
>>>
>>> My own search attempts coming up dry.   Just vaguely related academic
>>> material that doesn't even give me a good idea where to start!
>>>
>>> I approach the oracle!
>>>
>>> Mike
>>>
>>> _______________________________________________
>>> clue mailing list: clue at cluedenver.org
>>> For information, account preferences, or to unsubscribe see:
>>> http://cluedenver.org/mailman/listinfo/clue
>>>
>>
>>
>> _______________________________________________
>> clue mailing list: clue at cluedenver.org
>> For information, account preferences, or to unsubscribe see:
>> http://cluedenver.org/mailman/listinfo/clue
>>
>
>
> _______________________________________________
> clue mailing list: clue at cluedenver.org
> For information, account preferences, or to unsubscribe see:
> http://cluedenver.org/mailman/listinfo/clue
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://cluedenver.org/pipermail/clue/attachments/20130508/f6673385/attachment.html 


More information about the clue mailing list