Migrating to a new operating system

Migrating to Red Hat Enterprise Linux 9

For RHEL 9, the /tmp directory is a separate partition that has the nosuid, noexec, and nodev options set. If your apps install to and run scripts or binaries under the /tmp directory, they might fail. Update your apps to use the /var/tmp directory instead of the /tmp directory to run temporary scripts or binaries.

The default cgroup implementation is cgroup v2. In RHEL 9, cgroup v1 isn't supported. Review the Kubernetes migration documentation for cgroup v2 and verify that your applications fully support cgroup v2. There are known issues with older versions of Java that may cause out of memory (OOM) issues for workloads.

1. Review your worker pool operating systems to find which pools you need to migrate.

   ibmcloud ks worker-pools -c CLUSTER
   

2. Specify the REDHAT_9_64 version for the worker pool.

   ibmcloud oc worker-pool operating-system set --cluster CLUSTER --worker-pool POOL --operating-system REDHAT_9_64
   

3. Update each worker node in the worker pool by running the ibmcloud oc worker update for Classic clusters or ibmcloud oc worker replace for VPC clusters.

   Make sure you have enough worker nodes to support your workload while you update or replace the relevant worker nodes. For more information, see Updating VPC worker nodes or Updating classic worker nodes.

   Example command to update Classic worker nodes.

   ibmcloud oc worker update --cluster CLUSTER --worker WORKER1_ID [--worker WORKER2_ID] 
   

   Example command to replace VPC worker nodes.

   ibmcloud oc worker replace --cluster CLUSTER --worker WORKER_ID
   

4. Get the details for your worker pool and workers. In the output, verify that your worker nodes run the REDHAT_9_64 operating system.

   Get the details for a worker pool.

   ibmcloud oc worker-pools -c CLUSTER
   

   Get the details for a worker node.

   ibmcloud oc worker get --cluster CLUSTER --worker WORKER_NODE_ID