Uploaded image for project: 'OpenShift Bugs'
  1. OpenShift Bugs
  2. OCPBUGS-36859

OCP 4: primary iface selection by br-ex is flapping between reboots, hint files are ignored

XMLWordPrintable

    • Quality / Stability / Reliability
    • False
    • Hide

      None

      Show
      None
    • None
    • None
    • None
    • None
    • None
    • None
    • Done
    • Bug Fix
    • Hide
      Before this update, you could not override the node IP selection process for deployments that ran on installer-provisioned infrastructure. This limitation impacted user-managed loadbalancers that did not use VIP addresses on a machine network, and this caused problems in environments that had multiple IP addresses. With this release, deployments that run on installer-provisioned infrastructure now support the `NODEIP_HINT ` parameter for the `nodeip-configuration` systemd service. This support update ensures This support update ensures that the correct node IP address is used, even when the VIP addresses are not on the same subnet. (link:https://issues.redhat.com/browse/OCPBUGS-36859[OCPBUGS-36859])
      Show
      Before this update, you could not override the node IP selection process for deployments that ran on installer-provisioned infrastructure. This limitation impacted user-managed loadbalancers that did not use VIP addresses on a machine network, and this caused problems in environments that had multiple IP addresses. With this release, deployments that run on installer-provisioned infrastructure now support the `NODEIP_HINT ` parameter for the `nodeip-configuration` systemd service. This support update ensures This support update ensures that the correct node IP address is used, even when the VIP addresses are not on the same subnet. (link: https://issues.redhat.com/browse/OCPBUGS-36859 [ OCPBUGS-36859 ])
    • None
    • None
    • None
    • None

      Outcome Overview

      if we restart any worker node that is "Ready", on the first reboot CRIO fails to start, but on the second reboot CRIO starts successfully. 

      After restarting a node manually or when applying a MachineConfig, nodes are "NotReady" when the 'Wired connection 1' connection is associated with the "ens224" interface. When created in the "ens192" interface, the node becomes Ready.

      Logs (journalclt) when the node is "NotReady":

      NetworkManager[1296]: <info> [1719862141.3651] settings: (ens224): created default wired connection 'Wired connection 1'
      NetworkManager[1296]: <info> [1719862142.9642] policy: set 'Wired connection 1' (ens224) as default for IPv6 routing and DNS

       

      Logs (journalclt) when the node is "​Ready":

      NetworkManager[1284]: <info> [1719864127.3098] settings: (ens192): created default wired connection 'Wired connection 1'
      NetworkManager[1284]: <info> [1719864127.3624] policy: set 'Wired connection 1' (ens192) as default for IPv4 routing and DNS
      NetworkManager[1284]: <info> [1719864129.0021] policy: set 'Wired connection 1' (ens192) as default for IPv6 routing and DNS

       

      With each reboot /run/nodeip-configuration/primary-ip changes. Switches between IPv4 and IPv6. When it is IPv6 we have failure to start CRIO. The file /etc/systemd/system/kubelet.service.d/20-nodenet.conf always shows the 2 addresses.

              bnemec@redhat.com Benjamin Nemec
              rhn-support-hsahoo Himanshu Sekhar Sahoo
              None
              None
              Ross Brattain Ross Brattain
              Darragh Fitzmaurice Darragh Fitzmaurice
              Votes:
              5 Vote for this issue
              Watchers:
              14 Start watching this issue

                Created:
                Updated: