hjohanns

EPYC support on Ubuntu 18.04

Discussion created by hjohanns on Sep 6, 2018

I seem to have instability on a system Ubuntu 18.04 on EPYC 7351P

The MOBO is SuperMicro H11SSL.

I have installed the latest GA Ubuntu 18.04.

Even though many boards were saying the EPYC is not vulnerable to the same defects as the Ryzen, I am experiencing unstabilities.

The server, when left idle, will log strange messages.

Yesterday, the computer progressively crashed because I tried actions from apt and apt-get.

 

Let me ask again: are the EPYC processors vulnerable to the weakness that seem to affect the Ryzen? Can someone speak for this case?

Thanks a lot, hj

 

================Update 2019-09===================

Let me put this in context.

First I thank all the commenters that took the time to read this and ask for more questions. It is very appreciated.

Second currently the problem is solved / circumvented, in that it does not reproduce anymore.

In my analysis, I spotted first a problem logging a

Day 1 00:50:27 hostname1 kernel: [2099837.618882] general protection fault: 0000 [#1] SMP NOPTI

that you can find in the attached paste1.log.

Then pursuing the analysis, I had regular crashes related to the execution of fstrim on a SSD disk, which is a known issue, see for instance

systemd - nvme fstrim causing crash on linux, disabling with systemctl doesn't help - Unix & Linux Stack Exchange 

that would block the kernel in an uninterruptible state. See the attached fstrim log file.

I circumvented the problem by deactivating the regular fstrim. I know it is not ideal, but it was too difficult for me to investigate.

Since it did not reproduce since then, I guess that was the culprit.

So this has nothing to do with AMD or Ryzen, as I could see the issue happened also on Intel processors.

I have not recently checked on the web if the issue is resolved in recent kernels.

 

To conclude, I believe we can close the thread.

Thank you again for your suggestions and comments.

hjohanns

Outcomes