Uploaded image for project: 'Fast Datapath Product'
  1. Fast Datapath Product
  2. FDP-1724

CLONE [ovn25.03 fast-datapath-rhel-9] - ovn-nbctl --wait=sb returns too early

    • Icon: Bug Bug
    • Resolution: Done
    • Icon: Undefined Undefined
    • None
    • None
    • ovn25.03
    • 1
    • False
    • Hide

      None

      Show
      None
    • False
    • ovn25.03-25.03.1-70.el9fdp
    • rhel-9
    • None
    • rhel-net-ovn
    • ssg_networking

      The ovn-nbctl --wait=sb command could under certain conditions return earlier than the actual update is processed. The issue was discovered during review of https://patchwork.ozlabs.org/project/ovn/patch/20250915131602.54821-1-lucas.vdias@luizalabs.com/ the test would fail without the fix because northd entered commit fail loop. The expected behavior would be for the sync call to block until the timeout runs out. However the sync call was returning 0 and the test failed later on database check.

       

      The root cause is that northd I-P wouldn't run, but there was still valid transaction for SB. The transaction commit would return UNCHANGED, however, UNCHANGED moves idl loop next_cfg to cur_cfg, which is used by the synchronization. The next northd loop would update the SB nb_cfg in a different transaction than the failing one. Resulting in sync returning early.

              amusil@redhat.com Ales Musil
              ovnteam@redhat.com OVN Team
              Votes:
              0 Vote for this issue
              Watchers:
              2 Start watching this issue

                Created:
                Updated:
                Resolved: