Can you post the code snippet?  And the OneFS release you're on?

Hello, yes. I was going to, but had trouble with pasting into the forums for some reason.
So it's hosted here:

http://pastebin.com/qBrUkxAZ

It's OneFS 8.0.1

It's triggered by running the code on an nfs mount, and then (somewhere else that has it mounted - e.g. the isilon console will do)

#!/bin/sh
while true
do
   stat test.dat
done

I've also tried a packet capture of the above - there's a response to 'GETATTR' from the NFS server, which gives two different timestamps on the file - so it looks like it's _not_ client workstation specific.

Hi sobrique,

This looks similar to an issue seen prior to 7.2, but seems to be of a different variety. Either way, setting the time delta on the export may provide some relief.

Here is the knowledge base article about the older issue:

https://support.emc.com/kb/457853

Page 442 of the OneFS 8.0.0 CLI Administration Guide provides the commands for changing and reverting the time delta on an export:
https://support.emc.com/docu65065_OneFS-8.0.0-CLI-Administration-Guide.pdf?language=en_US

Yes, we've tried setting --time-delta to a lower number. This doesn't solve the issue, but as might be expected - does make it less frequent, as it only occurs when you're on an appropriate boundary.

isi exports modify --zone --time-delta 1

This sets the resolution to 1s, and means it's considerably less frequent.

time mismatch: 1469087524.000000000 != 1469087523.000000000

It serves as a workaround in the short term though.

Hi sobrique,

How long does this take to reproduce in your environment?  I've been running your steps for many hours without hitting the problem, and just want an expectation of how long I might need to wait.

> Without running the "stat" loop on another host, it can take a while.

> With it, it's seconds (if that).

>

> Which implies you aren't getting it.  Which OneFS version are you on? (we are 8.0.1)

Here's my setup:

• 3 node Isilon x410 cluster running OneFS 8.0.0.1.
• Two Linux clients, both CentOS 6.2.
• Both clients mount Isilon cluster using over 10G, using mount options "nfsvers=3,tcp". Both clients mount the same Isilon node.
• Client 1, cd /mnt/test/testdir and run the compiled code you provided.
• Client 2, cd /mnt/test/testdir and run the stat loop you posted above.

I haven't seen a time mismatch yet, but i'll let it keep running.

To clarify the release... you're saying this is 8.0.1, not 8.0.0.1?  8.0.1 hasn't been released yet and is only in Beta.  8.0.0.1 is currently the most recent GA code that we have.