Eldad Assis

Not that I know off the top of my head. Will probably require an extra step of preparing the values.yaml from a template where you inject values from env vars into it (with something like sed).

Or, you can use the helm upgrade --install <name> --set <key>=${IMAGE_NAME} option that will override the relevant default value of <key> with the ${IMAGE_NAME} value.

You should consider the helm chart just like any other versioned component.

Just like versioning of the Docker image or the binaries you build.

So just like I set a Docker image unique tag (version), I also set a helm chart version when I package it.

Be aware that helm version is expected to be semVer where as Docker tag can be any string.

But this is a single java process serving two services (on two ports). I want to run two separate tomcats running side by side (two java processes) based on the same classpath. This way I can harness the power of CDS (https://simonis.github.io/cl4cds/#_class_data_sharing)

Yes. I'd like to try... I guess passing a custom sever.xml and custom logging.properties should allow this.

Can it be done?

Hi Nassir. Great post.

Do you have an idea on how I can use the same paths exacly (single Tomcat dir) and spin up multiple instances? This will help me utilize the Java 10 Application Class-Data Sharing feature.

All JVM processes have to have an identical classpath to get it to work.

If running Jenkins master outside of K8s cluster, use the plugin. It covers configuration pains very well.

Correct. You are also welocome to contribute a pull request with support for such a configuration.

This is acually very easy! Since the pod by default has access to the kubernetes API, the kubectl picks up the local configuration and had access to API.

Just try opening a shell to a running pod, install kubectl and run "kubectl cluster-info". It works! Helm pics up this config aswell.

Note - in a hardened security, you might need to create and assign a strong service account this pod, but it should still work.