OpenShift Administrator’s Office Hour - Sizing OpenShift Clusters and Nodes

This week’s OpenShift Administrator’s Office Hour was dedicated to sizing. More specifically, we had a conversation about how to determine the appropriate size and count of worker nodes based on Pod requirements while also taking into account factors such as horizontal vs vertical scaling, failure domains, and how to avoid stranding resources.  

We had some really great viewer questions about this topic, including:

• Sizing considerations for 3-node “compact” clusters?
• Should I use swap or virtual RAM to increase density?
• Can I oversubscribe CPU/memory resources for OpenShift nodes?
• Is it better to scale out or scale up nodes?

See the list below for additional links to specific topics, questions, and supporting materials for the episode!

Please subscribe to the OpenShift.tv streaming calendar to see the upcoming episode topics and to receive any schedule changes. If you have questions or topic suggestions for the OpenShift Administrator’s Office Hour, please contact us via Discord, Twitter, or come join us live, Wednesdays at 11am EST / 1600 UTC, on YouTube and Twitch.

Thank you for watching!

Episode 15 recorded stream:

Supporting links for today’s topic:

• Episode 10’s topic was “Storage for Nodes”, which covered the workloads and considerations when sizing the disks used for OpenShift control plane and worker nodes.
• The start of the sizing topic is here. If you’re interested in the other questions discussed, please see the links in the next section.
• The “Recommended host practices” documentation page provides sizing guidance for control plane nodes and infrastructure nodes.

Other links and materials referenced during the stream:

• During the stream we reminded the audience of the recently announced RHSB-2021-002 and CVE-2021-3156. Please remember, it’s important to keep your OpenShift clusters updated, as Red Hat Enterprise Linux CoreOS (RHCOS) is patched alongside the other OpenShift services, features, and functions.
• We talked briefly about how RHCOS nodes configure their hostname, discussing how the node-valid-hostname service is used to set the hostname, but the value it uses can be set from multiple sources: virtual machine name, reverse DNS, DHCP assigned, and configuration from Ignition.

Other questions answered during the stream:

• Are there plans to create PowerShell modules for OpenShift? Unfortunately, I have not been able to find any RFEs regarding this feature, however there are some interesting-looking modules in the PowerShell Gallery for Kubernetes, which will also work with OpenShift.
• When deploying to virtual infrastructure, such as RHV or vSphere, can I control which hypervisor nodes the OpenShift virtual machines are scheduled on?
• One of our viewers asked about the conclusion of the disconnected OLM live stream.  You can find that stream (episode 13) here. If there are any questions or things we missed, please reach out using the contact info at the top of the post!