Create Dockerfile + anything else to deploy Forkana changes

[taoeffect]

Problem

Currently we have to manually setup and run Forkana locally to review changes.

I'd be useful if there were some private deployment site that was easily updateable when new PRs are merged so that we can review those changes more easily.

It's especially helpful for designers to see how it actually works.

Solution

Not entirely sure, but it would at least involve a Dockerfile that can be rebuilt to grab the latest changes from master and then redeployed.

[pedrogaudencio]

I had a look at the requirements and put together what we'd need for a simple dev environment that updates automatically on merges to master (or new release tags):

1. A Docker image we can build reliably

• Prefer the rootless image for a safer default (dev env, least privilege)
• Build should follow the existing repo build flow (frontend + backend + vendor) so it matches what we have locally

2. CI pipeline triggered by merges to master

• GitHub Actions: build → push image to a private registry → deploy
• Use concurrency to avoid overlapping deploys

3. A simple dev host

• A single VM running Docker Compose is enough for a first version
• Run Forkana + Postgres (and ideally Redis so it's closer to production behaviour)
• Basic dev-safe config: lock install, disable registration, set ROOT_URL, etc (current env-driven config already fits this)

4. Secrets + access

• Registry credentials (or GHCR permissions), deploy SSH key/host, app secret, db credentials

What this gets us

• A url (e.g dev.forkana.com) that always tracks the latest merged changes
• Much easier review loop for UI/UX/design changes
• Easy rollback via image tags (e.g. deploy sha-… if something breaks)

[taoeffect]

Nice, that sounds good to me!

A single VM running Docker Compose is enough for a first version

Agreed! 👍

Run Forkana + Postgres (and ideally Redis so it's closer to production behaviour)

Didn't realize Redis is used! Is Redis an add-on to Postgres, or is it used totally in place of it? Redis is a key-value database, not a relational database, so that's why I'm surprised.

[pedrogaudencio]

Didn't realize Redis is used! Is Redis an add-on to Postgres, or is it used totally in place of it? Redis is a key-value database, not a relational database, so that's why I'm surprised.

Redis would be an add-on to Postgres - this is mainly used for caching actually (expensive git operations e.g. last commit, fork contributor stats, commit counts per branch/tag) but also session storage, task queues and global locks. Postgres would be the main but there are several gains from having Redis as a cache layer.

For a single instance we can use the default (memory for cache, sessions and global locks and LevelDB for webhooks, code indexing, push updates and actions job scheduling), understanding that it has a few limitations (e.g. lost on restart/new deployments, but fine for the dev environment).

If this is a limitation for now, we can surely skip Redis while monitoring Postgres and memory usage.

[eighttrigrams]

Agreed with a docker-compose setup, and with skipping Redis for the beginning