Adding a Helm chart for deployment to Kubernetes

[jansenfuller]

I throw everything into Kubernetes and wanted a way to deploy TaskChampion to my cluster. I have tested this on my own cluster using TaskChamp as the client with great results
The chart has the following features:

• Support for SQLite and Postgres storage options
• Multiple replicas (only with Postgres as multiple replicas with SQLite would end with corrupting the database
• Custom PV location for SQLite database (defaults to a temp directory if not specified)
• Ingress and HTTPRoute configuration built in
• Client IDs stored in a Secret to prevent clients from being leaked
• Adjust envars to your liking

[KingOfSpades]

I was just about to build this. Will deploy tomorrow to my internal Talos cluster for testing.

[jansenfuller]

There is a weird issue where in the logs it shows the "starting entrypoint script..." multiple times. I'm unsure if that means it's unhealthy or its just how the application runs.

[KingOfSpades]

Im currently running sync server as a single container in Portainer and I don't see that behavior there. I'm going to test your chart tomorrow and report back

[KingOfSpades]

Looks good on my end, single deployment (sqlite) version

Side note, the secret ref name is my-generic-secret or taskchampion-client-ids?

[jansenfuller]

@KingOfSpades, Sorry I've been trying to get the Postgres implementation working correctly, it should be good to test if you like.

I'm unsure of what you're asking. The secret name can be whatever you want it to be. What is important is the key for client ids should be client-ids, and that should be a list of UUIDs.

If you would like to test out the Postgres implementation, you will need to manually put in your UUIDs into the database. I'm almost done updating the init container so that UUIDS are automatically added to the Postgres database after the migration has been ran.

[jansenfuller]

Auto add to Postgres on start is working now.

[jansenfuller]

Any input from @djmitche, @vagos?

[vagos]

Hey @jansenfuller thank you for sending this! Excuse my ignorance, but why does this need to live alongside the server source code? Could this be placed in a separate repository?

It might not be long-term viable to have configurations for all different kind of deployment frameworks (Ansible, Puppet, etc...).

Let me know if I am missing something, though.

[jansenfuller]

No that's a valid point. Since this is smaller project I felt it could go along side but if someone would like to put this into another repository, say helm-charts (so that other helm charts could be added in the future), that would work great. If I remember correctly, Helm best practices suggest putting charts in another repository. The "one repository = one artifact" pattern. Who would be in charge of that?

[djmitche]

I think this is fine to have in this repo.

Ideally it'd be somehow tested to make sure it doesn't get out of date, but I can imagine that's difficult to do in a way that would realistically catch a likely error. For exmaple, we might add a new env var configuration and forget to update the helm charts.

[jansenfuller]

Luckily there are options for users to add their own envars if we're missing one. There should be a env value where those can be added. I think the only real risk would be for things like multiple replicas, if metrics were to be exposed, etc. at that point there would be bigger things to worry about than the helm chart

[djmitche]

Sounds good. This is a great addition!

[jansenfuller]

@djmitche, when you have a moment, please publish the helm chart, it would probably be best to host it wherever the container is hosted

[djmitche]

I'm not familiar with Helm. Last I used it, there was some kind of helm publish. Do the charts go to image registries instead, now?

[jansenfuller]

Looks like there's a way to do it with a github action. It involves deploying it using GitHub Pages, which I unfortunately have no experience (I used Forgejo to test the chart). Here's a article that seems pretty straight forward to get it deployed. You can set the action

[jansenfuller]

I created a couple PRs that should get us started. One to publish the first version of the helm chart, and another to add a github action for every push after that.
#198
#199