keycloak-template.yaml Invokes command that causes Keycloak to use H2

[BarryDrez]

Description

Attempting to create and start a pod for Keycloak within OpenShift for multi-user che. Using the file: che/deploy/openshift/templates/multi/keycloak-template.yaml.

The file invokes the command: /scripts/kc_realm_user.sh. When tracing through the steps of the command, we can see that the standalone.xml file for Keycloak is never modified and defaults to allow Keycloak to use the H2 database.

Reproduction Steps

Create a project in OpenShift.
Create a pod for Postgresql using postgres-template.yaml (not really necessary)
Create a pod for Keycloak using keycloak-template.yaml
Note that the Postgresql db is not used and that H2 is.

OS and version:
Any OS, Latest Version

[BarryDrez]

There is something strange when I create a pod from the image: jboss/keycloak-openshift:3.4.3.Final and run the command:
/usr/bin/start-keycloak.sh -Dkeycloak.migration.action=import -Dkeycloak.migration.provider=dir -Dkeycloak.migration.strategy=IGNORE_EXISTING -Dkeycloak.migration.dir=/opt/jboss/keycloak/realms/ -Djboss.bind.address=0.0.0.0
I get the error:
java.lang.IllegalStateException: Directory /opt/jboss/keycloak/realms doesn't exists

It works when running che standalone in multiuser mode.

Wondering if this is related.

[BarryDrez]

@eivantsov Wondering if you were able to reproduce this.

[BarryDrez]

@eivantsov Can I please get an update on this?

[ghost]

There's a PR in works #11761

[BarryDrez]

Thank you

[skabashnyuk]

@l0rd @slemeur can you triage this issue, I think it might be important to fix for 7.0. WDYT?

[slemeur]

Before adding it to the scope, I think it's important to review this issue and check if this is still accurate. The deployment on openshift is now handle through the operator.
I'm labelling it for platform team.

[skabashnyuk]

@mkuznyetsov could you take a look?

[mkuznyetsov]

yes, on Operator it works with correct database, although it uses different keycloak image and it's entrypoint

[skabashnyuk]

@slemeur so, basically this issue is linked to the installation way that is no longer relevant.

[l0rd]

@mkuznyetsov @skabashnyuk so we close right?

[skabashnyuk]

@mkuznyetsov @skabashnyuk so we close right?

+0
Maybe it supposes to happen at the same time with #13888