Ability to Configure Update Behavior for Kubernetes/OpenShift Resources

[nexus-Six]

Description

The current implementation of the Kubernetes/OpenShift extensions use createOrReplace to apply the resources generated in target/kubernetes/openshift.yml to OpenShift. This leads to the replacement of all these Openshift resources on every redeploy, including the ServiceAccount. This new ServiceAccount in turn triggers the creation of three Openshift Secrets. Over time, this leads to a vast amount of stale Secrets in the namespace.

There should be a way to control this update behavior such as choosing between Replace (oc replace ...) and Update (oc apply ...).

Reproducer:

• Clone https://github.com/codingFonnegra/getting-started
  - oc login and create an OCP project
• In /getting-started execute multiple times : mvn package -Dquarkus.kubernetes.deploy=true -Dquarkus.profile=kube -DskipTests -Dquarkus.openshift.route.expose=true
• You can see how 3 new Secrets are created on every deploy

Implementation ideas

Perhaps a new configuration such as

quarkus.openshift.resources.creation-strategy=replace|update

could be a solution

[quarkus-bot]

/cc @geoand, @iocanel

[geoand]

Also cc @manusa

[manusa]

We do have a flag like this on JKube.

Perhaps a new configuration such as

quarkus.openshift.resources.creation-strategy=replace|update

I like this idea, having a well set defined creation/deployment strategies and then performing the appropriate operation in the deployment phase sound like a great solution and would add flexibility for the users.

There's now the option to perform server-side apply replacements too (#3334).

• create: Create only -> fails if resource exists ?
• replace: Delete and then create
• server-side-apply
• ...

[geoand]

@nexus-Six would the proposal from @manusa to also include server-side-apply be beneficial for you?

[nexus-Six]

Yes, this looks like good solution Am 26. Juli 2022 06:57:30 UTC schrieb Georgios Andrianakis ***@***.***>: ***@***.*** would the proposal from @manusa work for you?

…

-- Reply to this email directly or view it on GitHub: #26789 (comment) You are receiving this because you were mentioned. Message ID: ***@***.***>

[geoand]

Was closed due to my erroneous linkage to a PR