Uploaded image for project: 'RHEL'
  1. RHEL
  2. RHEL-1428

Race when using DHCP and copy-mac-from

    • Major
    • sst_network_management
    • ssg_networking
    • 22
    • 3
    • False
    • Hide

      None

      Show
      None
    • No
    • NMT - RHEL 8.10/9.4 DTM 18-20
    • Hide

      Given a system administrator configuring a node and applying a NMState ovs-bridge config with copy-from-mac set to the underlying physical interface,
      When the system administrator reboot the node, 
      Then the node should always request a DHCP address using the MAC address that the ovs-bridge was configured to copy. 

      Definition of Done

      • The implementation meets the acceptance criteria
      • The unit tests and integration tests are written and passed
      • The code is part of a build attached to an errata
      Show
      Given a system administrator configuring a node and applying a NMState ovs-bridge config with copy-from-mac set to the underlying physical interface, When the system administrator reboot the node,  Then the node should always request a DHCP address using the MAC address that the ovs-bridge was configured to copy.  Definition of Done The implementation meets the acceptance criteria The unit tests and integration tests are written and passed The code is part of a build attached to an errata
    • If docs needed, set a value
    • Unspecified

      Description of problem: There seems to be a problem when creating an ovs-bridge with copy-mac-from and assigning its address with DHCP. After rebooting, some of my nodes are requesting an address using a different MAC from the one the bridge was supposed to copy. Others correctly use the copied MAC. In both cases the MAC was set correctly, but it appears there is a timing issue with the DHCP request where sometimes the MAC is set after the request has been sent.

      I'm not sure whether this is an NMState issue or a NetworkManager one, but since I can't figure out where NMState is overriding the MAC it's difficult to provide a reproducer for NetworkManager in isolation.

      Version-Release number of selected component (if applicable): 2.2.13

      How reproducible: About 50% of the time in my experience.

      Steps to Reproduce:
      1. Apply NMState ovs-bridge config with copy-from-mac set to the underlying physical interface.
      2. Reboot node

      Actual results: 50% of the nodes will request a DHCP address using the wrong MAC address and get a different IP than they did when NMState initially created the bridge.

      Expected results: Nodes all get the same address they had before the reboot.

      Additional info: I will attach the NMState YAML and NM trace logs for both a node that did get the same address as well as one that demonstrates the problem.

            fge@redhat.com Gris Ge
            bnemec@redhat.com Benjamin Nemec
            Gris Ge Gris Ge
            Mingyu Shi Mingyu Shi
            Votes:
            0 Vote for this issue
            Watchers:
            12 Start watching this issue

              Created:
              Updated:
              Resolved: