[gpfsug-discuss] How to find which node is generating high iops in a GPFS 3.5

[Oesterlin, Robert]

You do realize 3.5 is out of service, correct? You should be looking at upgrading :-)

Catching this is real time, when you have a large number of nodes is going to be tough. How you recognizing that the file system is overloaded? Waiters? Looking at which nodes/NSDs have the longest/largest waiters may provide a clue.

You might also take a look at mmpmon – it’s a bit difficult to use in its raw state, but it does provide some good stats on a per file system basis. But you need to track these over times to get what you need.

Bob Oesterlin
Sr Principal Storage Engineer, Nuance


From: <gpfsug-discuss-bounces at spectrumscale.org> on behalf of Richard Lefebvre <richard.lefebvre+gpfsug at calculquebec.ca>
Reply-To: gpfsug main discussion list <gpfsug-discuss at spectrumscale.org>
Date: Monday, September 18, 2017 at 2:18 PM
To: gpfsug <gpfsug-discuss at spectrumscale.org>
Subject: [EXTERNAL] [gpfsug-discuss] How to find which node is generating high iops in a GPFS 3.5

Hi I have a 3.5 GPFS system with 700+ nodes. I sometime have nodes that generate a lot of iops on the large file system but I cannot find the right tool to find which node is the source. I'm guessing under 4.2.X, there are now easy tools, but what can be done under GPFS 3.5.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://gpfsug.org/pipermail/gpfsug-discuss_gpfsug.org/attachments/20170918/3862d74b/attachment-0002.htm>