OpenShift 5.0 on IBM Power: Using rhel-9 or rhel-10 OS Image Streams

This blog was originally posted to https://community.ibm.com/community/user/blogs/paul-bastide/2026/11/12/openshift-5-0-on-ibm-power-using-rhel-9-or-rhel-10-os-image-streams


For an existing OpenShift cluster running the rhel-9 OS image stream, upgrading the OpenShift platform from 4.x to 5.0 does not itself move the node operating system to rhel-10. The platform upgrade and the node OS stream change are separate steps, so the cluster remains on rhel-9 unless an administrator deliberately changes the stream.

That distinction is essential for Power9 customers: the rhel-10 stream is not supported on Power9. Keep Power9 clusters on rhel-9; do not apply the switch procedure below to a cluster that includes Power9 nodes.

Note: If you are installing a fresh cluster, it defaults to rhel-10. You can configure your install-config.yaml to set osImageStream: rhel-9. We have updated ocp4-upi-powervm and ocp4-upi-powervs so you can set os_image_stream to either rhel-9 or rhel-10.

Before considering a move to rhel-10

Only plan a move to rhel-10 after confirming all of the following:

  • The cluster is updated to OpenShift 5.0.
  • Every node uses hardware compatible with RHCOS 10. Power9 is not supported by the rhel-10 stream.
  • You can move all intended nodes to RHCOS 10. A cluster with a mixture of RHCOS 9.x and RHCOS 10.x nodes is not supported.

If any machine config pools have an explicit OS stream override, review them as part of the plan. Changing the cluster-wide default affects pools that inherit that default; do not leave the cluster in a mixed RHCOS 9.x/RHCOS 10.x state.

Check that the rhel-10 stream is available

Run the following command and confirm that rhel-10 appears in the output:

oc get osImageStreams/cluster -o yaml | grep rhel-10

Expected output includes:

- name: rhel-10

Change the cluster-wide default stream

For a cluster on hardware supported by RHCOS 10, set the OSImageStream default to rhel-10:

oc patch osimagestream cluster --type merge -p '{"spec":{"defaultStream":"rhel-10"}}'

This changes the cluster-wide default and starts the node OS rollout for pools that use that default. Monitor the machine config pools and wait for the rollout to complete. Do not use this command on Power9 systems.

Note: OpenShift 5.0 does not support a mixture of RHCOS 9.x and 10.x nodes; requires all nodes to move to RHCOS 10.x. So changing just one MachineConfigPool while the others remain on rhel-9 is not a supported steady state. For the documented migration, use the cluster-wide OSImageStream default change from the blog and let the pools complete the migration. If the target pool includes Power9 nodes, do not set it to rhel-10.

If you are patching the built-in worker pool, check custom pools too: pools without their own stream override can inherit the worker stream.

Verify the rollout

After the nodes in a pool return to the READY state, inspect its status and confirm the stream is rhel-10:

oc get machineconfigpool -n openshift-machine-config-operator <machine_config_pool_name> -o yaml

Look for:

status:
  osImageStream:
    name: rhel-10

You can also check the OS version reported by an individual node:

oc debug node/<node_name> -- chroot /host cat /etc/redhat-release

The takeaway

An OpenShift 4.x-to-5.0 upgrade does not automatically change an existing rhel-9 node OS stream. For Power9, rhel-9 remains the supported stream; rhel-10 is not supported on Power9. Treat any move to rhel-10 as a separate, planned cluster-wide OS migration, and proceed only after confirming hardware compatibility and the required runtime and rollout prerequisites.

Further reading

Comments

Leave a Reply

Your email address will not be published. Required fields are marked *