RedHat Keycloak operator rolling update expectations

Configuring the server
1

Hi!

I’m running Keycloak in OpenShift through the RedHat operator: rhbk-operator.v24.0.10-opr.1

Now, I was expecting the operator to be able to perform rolling updates, in such a way that it will not kill all the pods in one go and drop all the user sessions. However, it appears that when I do certain manifest changes, such as tweaking environment variables or log levels, I get my desired behaviour. But when I want to bump the keycloak docker image, then that won’t deploy smoothly.

I’d like to set either set my expectations straight, or try to understand what I might be doing wrong.

Anyone?

Many thanks,
David

2

AFAIK, rolling updates are not something that is currently supported by Keycloak. While you can do them in some cases, the operator doesn’t seem to support this.

3

There is currently some work in progress for supporting zero downtime upgrades with operator, depending on the version. Eventually there will be release something in future versions, it won’t be anything “big bang”, there will be a step-by-step improvement of the upgrade experience with operator.

4

I found the issue again: Initial Zero-Downtime Image handling in KC Operator · Issue #36305 · keycloak/keycloak · GitHub