In this section: https://docs.redhat.com/en/documentation/red_hat_openstack_platform/17.1/html/framework_for_upgrades_16.2_to_17.1/upgrading-a-standard-overcloud_upgrading-overcloud-standard#upgrading-rhosp-on-all-nodes-in-each-stack_overcloud-upgrade

We often get asked 'can we batch this?' We need to improve this page to provide the answer and to help customers understand why this is important to do all at once, and what it does.

There is also the note 'If you need assistance, contact Red Hat Support.' which confuses everyone, including support. If we cannot provide a link to a KCS for an alternative option, then we shouldn't have that sentence.

Some helpful hints:

• During this step, only the OpenStack containers and packages are updated. No system upgrade happens (that happens in Chapters 9-11).
• The process has no data plane down time.
• As mentioned in Chapter 1.4, the OpenStack APIs are interrupted.
• The speed of execution will depend on multiple factors which are described in the introduction paragraph of Chapter 1.4.
• The execution of the software & service upgrades happens to multiple roles in parallel, but each role is executed sequentially. If there is a role with 100 nodes it will take longer to execute than a role with 50 nodes. However, the whole run will not complete until all roles are complete. The benchmark timing in Chapter 1.4 had 4 compute roles in it. One of the compute roles had 166 nodes in it.

It is highly recommended that the procedure is used as-is, without attempting to batch.

If a customer does attempt to batch the execution, please note that it is a complex process which requires deep understanding, planning, lab validation and execution. The following are the risks.

• If a customer using ML2/OVN attempts to batch it, there will be a data plane outage if the ordering and selected hosts executed together are not correct.
• If a customer attempts to batch the computes, the ssh keys used for live migrations will not be populated correctly. This will require additional actions after the upgrade to correct.