[gpfsug-discuss] Changing verbsPorts On Single Node

[Bryan Banister]

I agree with this assessment.  I would also recommend looking into user defined node classes so that your mmlsconfig output is more easily readable, otherwise each node will be listed in the mmlsconfig output.

HTH,
-Bryan

From: gpfsug-discuss-bounces at spectrumscale.org [mailto:gpfsug-discuss-bounces at spectrumscale.org] On Behalf Of David D. Johnson
Sent: Wednesday, February 22, 2017 10:12 AM
To: gpfsug main discussion list <gpfsug-discuss at spectrumscale.org>
Subject: Re: [gpfsug-discuss] Changing verbsPorts On Single Node

I have a feeling that this is how mmchconfig is supposed to work. You’ve asked it to change the
configuration of one node, but the database of configuration settings needs to be propagated to
the entire cluster whenever a change is made.  You’ll find a section in the mmlsconfig output specific
to the node(s) that have been changed [node155] …. At this point your configuration may be out of
sync on any number of nodes.

 — ddj
Dave Johnson
Brown University CCV/CIS

On Feb 22, 2017, at 10:57 AM, Douglas Duckworth <dod2014 at med.cornell.edu<mailto:dod2014 at med.cornell.edu>> wrote:

Hello!

I am an HPC admin at Weill Cornell Medicine in the Upper East Side of Manhattan.  It's a great place with researchers working in many computationally demanding fields.  I am asked to do many new things all of the time so it's never boring.  Yesterday we deployed a server that's intended to create atomic-level image of a ribosome.  Pretty serious science!
We have two DDN GridScaler GPFS clusters with around 3PB of storage.  FDR Infiniband provides the interconnect.  Our compute nodes are Dell PowerEdge 12/13G servers running Centos 6 and 7 while we're using SGE for scheduling.  Hopefully soon Slurm.  We also have some GPU servers from Pengiun Computing, with GTX 1080s, as well a new Ryft FPGA accelerator.  I am hoping our next round of computing power will come from AMD...

Anyway, I've been using Ansible to deploy our new GPFS nodes as well as build all other things we need at WCM.  I thought that this was complete.  However, apparently, the GPFS client's been trying RDMA over port mlx4_0/2 though we need to use mlx4_0/1!  Rather than running mmchconfig against the entire cluster, I have been trying it locally on the node that needs to be addressed.  For example:

sudo mmchconfig verbsPorts=mlx4_0/1 -i -N node155

When ran locally the desired change becomes permanent and we see RDMA active after restarting GPFS service on node.  Though mmchconfig still tries to run against all nodes in the cluster!  I kill it of course at the known_hosts step.

In addition I tried:

sudo mmchconfig verbsPorts=mlx4_0/1 -i -N node155 NodeClass=localhost

However the same result.

When doing capital "i" mmchconfig does attempt ssh with all nodes.  Yet the change does not persist after restarting GPFS.

So far I consulted the following documentation:

http://ibm.co/2mcjK3P
http://ibm.co/2lFSInH

Could anyone please help?

We're using GPFS client version 4.1.1-3 on Centos 6 nodes as well as 4.2.1-2 on those which are running Centos 7.

Thanks so much!

Best
Doug


Thanks,

Douglas Duckworth, MSc, LFCS
HPC System Administrator
Scientific Computing Unit
Physiology and Biophysics
Weill Cornell Medicine
E: doug at med.cornell.edu<mailto:doug at med.cornell.edu>
O: 212-746-6305
F: 212-746-8690
_______________________________________________
gpfsug-discuss mailing list
gpfsug-discuss at spectrumscale.org<http://spectrumscale.org>
http://gpfsug.org/mailman/listinfo/gpfsug-discuss


________________________________

Note: This email is for the confidential use of the named addressee(s) only and may contain proprietary, confidential or privileged information. If you are not the intended recipient, you are hereby notified that any review, dissemination or copying of this email is strictly prohibited, and to please notify the sender immediately and destroy this email and any attachments. Email transmission cannot be guaranteed to be secure or error-free. The Company, therefore, does not make any guarantees as to the completeness or accuracy of this email or any attachments. This email is for informational purposes only and does not constitute a recommendation, offer, request or solicitation of any kind to buy, sell, subscribe, redeem or perform any type of transaction of a financial product.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://gpfsug.org/pipermail/gpfsug-discuss_gpfsug.org/attachments/20170222/e6f497cd/attachment.htm>