Uploaded image for project: 'OpenShift Bugs'
  1. OpenShift Bugs
  2. OCPBUGS-11507

Potential 4.12 to 4.13 upgrade failure due to NIC rename

    XMLWordPrintable

Details

    • 5
    • Sprint 234 - Update&Remoting
    • 1
    • No
    • Proposed
    • False
    • Hide

      None

      Show
      None

    Description

      This is a clone of issue OCPBUGS-10787. The following is the description of the original issue:

      STATUS: We basically know the general shape of what we need to do, and PoC work exists to do it and is queued up in https://github.com/openshift/machine-config-operator/pull/3650 for testing.  However, uncertainty remains around:

       

      • Whether PoC code actually works e2e
      • The potential blast radius of the changes that could affect other scenarios (trying to minimize)

       

      Description of problem:

      Upgrades to from OpenShift 4.12 to 4.13 will also upgrade the underlying RHCOS from 8.6 to 9.2. As part of that the names of the network interfaces may change. For example `eno1` may be renamed to `eno1np0`. If a host is using NetworkManager configuration files that rely on those names then the host will fail to connect to the network when it boots after the upgrade. For example, if the host had static IP addresses assigned it will instead boot using IP addresses assigned via DHCP.

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

      4.13

      How reproducible:

      Always.

      Steps to Reproduce:

      1. Select hardware (or VMs) that will have different network interface names in RHCOS 8 and RHCOS 9, for example `eno1` in RHCOS 8 and `eno1np0` in RHCOS 9.
      
      1. Install a 4.12 cluster with static network configuration using the `interface-name` field of NetworkManager interface configuration files to match the configuration to the network interface.
      
      2. Upgrade the cluster to 4.13.
      

      Actual results:

      The NetworkManager configuration files are ignored because they don't longer match the NIC names. Instead the NICs get new IP addresses from DHCP.

      Expected results:

      The NetworkManager configuration files are updated as part of the upgrade to use the new NIC names.

      Additional info:

      Note this a hypothetical scenario. We have detected this potential problem in a slightly different scenario where we install a 4.13 cluster with the assisted installer. During the discovery phase we use RHCOS 8 and we generate the NetworkManager configuration files. Then we reboot into RHCOS 9, and the configuration files are ignored due to the change in the NICs. See MGMT-13970 for more details.

      Attachments

        Issue Links

          Activity

            People

              rhn-gps-dmabe Dusty Mabe
              openshift-crt-jira-prow OpenShift Prow Bot
              Joshua Clark Joshua Clark
              Gris Ge
              Votes:
              0 Vote for this issue
              Watchers:
              10 Start watching this issue

              Dates

                Created:
                Updated:
                Resolved: