1. Documentation
×

Preparing to perform an EUS-to-EUS update

Due to fundamental Kubernetes design, all OKD updates between minor versions must be serialized. You must update from OKD 4.8 to 4.9 and then to 4.10. You cannot update from OKD 4.8 to 4.10 directly. However, beginning with the update from OKD 4.8 to 4.9 to 4.10, administrators who wish to update between two Extended Update Support (EUS) versions can do so incurring only a single reboot of non-control plane hosts.

There are a number of caveats to consider when attempting an EUS-to-EUS update.

  • EUS-to-EUS updates are only offered after updates between all versions involved have been made available in stable channels.

  • If you encounter issues during or after upgrading to the odd-numbered minor version but before upgrading to the next even-numbered version, then remediation of those issues may require that non-control plane hosts complete the update to the odd-numbered version before moving forward.

  • You can complete the update process during multiple maintenance windows by pausing at intermediate steps. However, plan to complete the entire update within 60 days. This is critical to ensure that normal cluster automation processes are completed including those associated with certificate rotation.

  • You must be running at least OKD 4.8.14 before starting the EUS-to-EUS update procedure. If you do not meet this minimum requirement, update to a later 4.8.z before attempting the EUS-to-EUS update.

  • Support for RHEL7 workers was removed in OKD 4.10 and replaced with RHEL8 workers, therefore EUS-to-EUS updates are not available for clusters with RHEL7 workers.

  • Node components are not updated to OKD 4.9. Do not expect all features and bugs fixed in OKD 4.9 to be made available until you complete the update to OKD 4.10 and enable all MachineConfigPools to update.

  • All the clusters might update using EUS channels for a conventional update without pools paused, but only clusters with non control-plane MachineConfigPools objects can do EUS-to-EUS update with pools paused.

EUS-to-EUS update

The following procedure pauses all non-master MachineConfigPools and performs updates from OKD 4.8 to 4.9 to 4.10, then unpauses the previously paused MachineConfigPools. Following this procedure reduces the total update duration and the number of times worker nodes are restarted.

Prerequisites

  • Review the release notes for OKD 4.9 and 4.10

  • Review the release notes and product lifecycles for any layered products and OLM Operators. Some may require updates either before or during an EUS-to-EUS update.

  • Ensure that you are familiar with version-specific prerequisites, such as administrator acknowledgement that is required prior to updating from OKD 4.8 to 4.9.

  • Verify that your cluster is running OKD version 4.8.14 or later. If your cluster is running a version earlier than OKD 4.8.14, you must update to a later 4.8.z version before updating to 4.9. The update to 4.8.14 or later is necessary to fulfill the minimum version requirements that must be performed without pausing MachineConfigPools.

  • Verify that MachineConfigPools is unpaused.

Procedure

  1. Upgrade any OLM Operators to versions that are compatible with both versions you are updating to.

  2. Verify that all MachineConfigPools display a status of UPDATED and no MachineConfigPools display a status of UPDATING. View the status of all MachineConfigPools, run the following command:

    $ oc get mcp
    Example output

    Output is trimmed for clarity:

    NAME     CONFIG                                         	UPDATED   UPDATING
master   rendered-master-ecbb9582781c1091e1c9f19d50cf836c       True  	  False
worker   rendered-worker-00a3f0c68ae94e747193156b491553d5       True  	  False

  3. Pause the MachineConfigPools you wish to skip reboots on, run the following commands:

    You cannot pause the master pool.

    		 
    $ oc patch mcp/worker --type merge --patch '{"spec":{"paused":true}}'

  4. Change to the eus-4.10 channel, run the following command:

    $ oc adm upgrade channel eus-4.10

  5. Update to 4.9, run the following command:

    $ oc adm upgrade --to-latest
    Example output
    Updating to latest version 4.9.18

  6. Ensure the 4.9 updates are completed successfully retrieve the cluster version, run the following command:

    $ oc get clusterversion
    Example output
    NAME  	  VERSION  AVAILABLE  PROGRESSING   SINCE   STATUS
version   4.9.18   True       False         6m29s   Cluster version is 4.9.18

  7. If necessary, upgrade OLM Operators using the Administrator perspective on the web console.

  8. Update to 4.10, run the following command:

    $ oc adm upgrade --to-latest

  9. Ensure the 4.10 update is completed successfully retrieve the cluster version, run the following command:

    $ oc get clusterversion
    Example output
    NAME  	  VERSION  AVAILABLE  PROGRESSING   SINCE   STATUS
version   4.10.1   True       False         6m29s   Cluster version is 4.10.1

  10. Unpause all previously paused MachineConfigPools, run the following command:

    $ oc patch mcp/worker --type merge --patch '{"spec":{"paused":false}}'

    If pools are not unpaused, the cluster is not permitted to update to any future minors and maintenance tasks such as certificate rotation are inhibited. This puts the cluster at risk for future degradation.

  11. Verify that your previously paused pools have updated and your cluster completed the update to 4.10, run the following command:

    $ oc get mcp
    Example output

    Output is trimmed for clarity:

    NAME 	   CONFIG                                            UPDATED     UPDATING
master   rendered-master-52da4d2760807cb2b96a3402179a9a4c    True  	 False
worker   rendered-worker-4756f60eccae96fb9dcb4c392c69d497    True 	 False