Hi there,

I hope it's okay to ask this question here. I've been running some MPI programs on our cluster and got some errors that indicated something wrong with the fabric. I then ran ibdiagnet to see if I could pinpoint the issue, but then received a complaint from sys admin that doing so could negatively impact the cluster performance. I was wondering how much impact this could possibly have? The entire diagnostic took around 8 minutes to run -- which means it's not spending much time on each compute node. Could it impact the switches badly? I can't imagine it uses much of the bandwidth? I've asked and they couldn't elaborate why running the diagnostic may be harmful -- does someone here know perhaps?


Post has attachment
Email:  sixshy@gmail.com
5 Photos - View album

Post has attachment
Bridgeworks protects organisations’ existing investments in iSCSI servers and Fibre Channel devices | http://bit.ly/YEfUfe

Post has attachment
Thanks so much for Joining and help build the Mellanox Community. In the effort of continuing to grow the Mellanox Community program we have now opened our own community. Be one of the first to sign up and connect with Mellanox.

Happy New Year. Want a change to Win, Take a pic wearing a Mellanox shirt and post to Google Plus. We will send the best pics some gift cards

Welcome to the new Mellanox Community. If you have any questions let us know.
Wait while more posts are being loaded