Uploaded image for project: 'RHEL'
  1. RHEL
  2. RHEL-30113

KSM doesn't merge duplicated data on multi-node NUMA systems

    • Icon: Bug Bug
    • Resolution: Unresolved
    • Icon: Undefined Undefined
    • None
    • rhel-9.5
    • ksmtuned
    • None
    • Moderate
    • None
    • QE ack
    • False
    • Hide

      None

      Show
      None
    • None
    • Red Hat Enterprise Linux
    • None
    • x86_64
    • Linux
    • None

      What were you trying to do that didn't work?
      KSM is only merging a small portion of duplicated data, allocated at the same time with several workers in the VM. It's important to say that this only happens in RHEL 9 (at least the last three releases) and when the system has more than one NUMA node. It's not depending in the amount of memory the system has and THP is being disabled during the test execution.

      Please provide the package NVR for which bug is seen:
      Any RHEL 9, at least from the last three minors, e.g. RHEL 9.5
      Any QEMU version, e.g. qemu-kvm-8.2.0-6.el9

      How reproducible:
      100%

      Steps to reproduce
      1. Run any ksm_overcommit avocado test in a RHEL 9 host with more than one NUMA node.
      2. python3 ConfigTest.py --guestname=RHEL.9.5.0 --testcase=ksm_overcommit.ksm_parallel
      3. Check the case failed because only a portion of the data has been merged

      Expected results
      KSM to deduplicate almost all the memory.

      Actual results
      KSM is only merging a fraction, not enough, of the total duplicated memory.

        1. debug.log
          105 kB
          Mario Casquero

              dhildenb@redhat.com David Hildenbrand
              mcasquer@redhat.com Mario Casquero
              David Hildenbrand David Hildenbrand
              Mario Casquero Mario Casquero
              Votes:
              0 Vote for this issue
              Watchers:
              8 Start watching this issue

                Created:
                Updated: