SLES12SP4 and upper version

We recently encountered a SLES12SP4, or SLES15SP1 system crash (soft lockup...) when running our applications. After some trial and error, we found that we have to disable numa_balancing in kernel settings to resolve the problem. NUMA balancing is good feature in OS that is to boost the normal application performance by ~20%, however, due to the "bug", we can not enable it.
Does anybody have the same problem? I am wondering if it is a SLES OS bug that SUSE has already known and has a fix on it.

Comments

  • malcolmlewismalcolmlewis Knowledge Partner

    Hi and welcome to the forum :)
    Are you running btrfs? If so can you try the following boot option: cgroup_disable=memory

  • Thanks for the suggestion, I believe we have btrfs running. I have not made the boot option test yet, but even if the setting is working to make the system sustainable, I still need to understand what's the problem that actually caused the system crash. Is there a known bug in btrfs or in the NUMA balancing implementation, or a combination of these or others?

  • malcolmlewismalcolmlewis Knowledge Partner

    Hi
    Are you in a position to open a Support Request about this?

  • yes, if I can not find resolution from the forum, I may submit a support request. Thanks.

  • malcolmlewismalcolmlewis Knowledge Partner

    Hi
    A Support Request might be your best course of action, I suspect you need to work with the engineers and test potential fixes.

Sign In or Register to comment.