setup pnpm and Dockerfile

[jonathan-irvin]

• What kind of change does this PR introduce? (Bug fix, feature, docs update, ...)
  We need a good production Dockerfile. Dev can come later as the dev environment itself will need to be handled carefully, but sans-Docker, pnpm is a great candidate for establishing some standards in our environments.

Using pnpm for package management over npm

For one, this enforces the node version to be 20.9.0
That's right, no matter what command you use, pnpm will make sure its on the right version of docker.

• Why was this change needed? (You can also link to an open issue here)
  One of our chief problems is "it works on my machine".
  I aim to fix that.

• Aligns package versions using pnpm which is great for maintaining package integrity with it's lockfile system:

See: https://pnpm.io/motivation & https://pnpm.io/feature-comparison

• Sets up a Dockerfile (WIP) for building the Production image of postiz-app using a multi-stage build.

This will need CI and also some other things that are still missing in deps since those items will need to be added to CLI (like Python which I added in Docker)

• Other information:

[vercel]

@jonathan-irvin is attempting to deploy a commit to the Listinai Team on Vercel.

A member of the Team first needs to authorize it.

[nevo-david]

Hi! Did you check this?
I think this command will run in an endless loop.
Will try it tomorrow :)

[jonathan-irvin]

It won't run in an endless loop because it doesn't work :)
I'm having trouble with some missing dependencies again. In lieu of building all three, I was attempting to isolate the frontend.

The end-goal is to build them all separately into 3 different container images so they can be hosted separately for kubernetes...I just can't get them to build.

Any tips would be astronomical!

[jamesread]

We need to agree consistency between these difference dockerfiles, because if one is on Alpine, another on Fedora, one installs python and stuff, and the other doesn't, it'll create a support mess.

I'm not saying which approach is better or worse, not at all, yours might be the way this goes - just we need to have a conversation before either PR gets merged about which is the technically better approach, so they're both consistent.

[jonathan-irvin]

Yeah, I'm starting a baseline from barebones and assuming nothing is in place and adding as we go. We can refine from there.

From @nevo-david, I know you like Fedora, I respect that, but maybe you can maintain a separate fork with a fedora container image?

Can you setup Github actions and prove it out?

https://github.com/jonathan-irvin/postiz-app/actions/runs/10753709803/job/29823454793

[jonathan-irvin]

Our "problem" @jamesread is that we don't know the complete picture of dependencies. I added python because node-gyp needs it to run as well as some other things to build the deps on the arch of choice.

Still some setup stuff missing.

I just want it to build successfully first. Then dev images and scaling out infrastructure comes later.

[jamesread]

Hey @jonathan-irvin , I was also going to setup GitHub actions as well (was talking about this on Discord), but I saw you did it, so didn't want to "step on your toes". I'm more than happy to have a play if that won't annoy you?

Personally I can build just fine with the default repo, so I'm not sure what's going wrong for you. Obviously reproducing things in a clean CI environment is a great way to go. Node v20.12.2 here.

I want to be clear that I chose Fedora because technically it's a better solution - I've run projects before where people requested things that node/npm images don't ship, and so it provides options for the future. I made those comments in the other issue about why I think Fedora would be a good choice, but I think we should chose based on best technical reasons, not feelings, so if we come up with reasons why alpine is technically better, let's do that.

[jamesread]

It took me about 15 minutes, but I have a clean build and working GitHub action - let's chat in #208 .

[nevo-david]

One more thing, look at this thread on X:
https://x.com/nevodavid/status/1830513523014987954

I also thought, like you, that if you run nx run-many with production variables and build it runs it for a production env.
But it doesn't.
When running the project, make sure you run it with node dir.
This is why I created
"start:prod"
"start:prod:frontend"
"start:prod:workers"
"start:prod:cron"

The only one that can start with the next command is the front end.
you will see memory drops from 3.4gb to 135mb

[jamesread]

@jonathan-irvin as per recent discussions, does it make sense to close this in favour of us jointly working on #205 ?