Skip to Content.
Sympa Menu

perfsonar-user - Re: [perfsonar-user] PS 3.51. and MLX 100g card, problems

Subject: perfSONAR User Q&A and Other Discussion

List archive

Re: [perfsonar-user] PS 3.51. and MLX 100g card, problems


Chronological Thread 
  • From: Phil Reese <>
  • To:
  • Subject: Re: [perfsonar-user] PS 3.51. and MLX 100g card, problems
  • Date: Fri, 10 Mar 2017 16:04:44 -0800
  • Ironport-phdr: 9a23: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

Looks like a colleague here made some progress, he writes:
Found a process "perfsonar-configure_nic_parameters" which caused the kernel
panic. The service is currently disabled.

At least it can be massaged into operation now that it is functional.

Hope this may help someone else!

Phil



On 3/10/17 12:27 PM, Phil Reese wrote:
Hi,

We're trying to get the current production version of PS installed on a Dell. We want to make this a 100g tester box, so have a 100g Mellanox MCX456A-ECA card installed.

After a normal install of PS, just before the system would come up, we consistently get the following error:

Loading mlx5_core.ko module
Out of memory: Kill process 231 (insmod) score 1 or sacrifice child
Killed process 231, UID 0, (insmod) total-vm:1424kV, anon-rss: 320kB, file-rss:12kB

And the system is hung, only a power cycle brings it back.

Moving to single user, we can get on the system but even moving all the mlx* drivers out of the way, the error still happens.

We're pretty sure there is an easy fix but wanted to ask this group before investing more time into the effort. Really would like to see PS4 and Centos 7 as soon a possible.

Thanks,
Phil









Archive powered by MHonArc 2.6.19.

Top of Page