Running Multiple Rails Apps Concurrently with Foreman and Procfile.dev

Pizza

Pizza courtesy of Pizza for Ukraine!

As I've said, I build a lot of side projects and I really, really like the model of having:

ALL MY APPS RUNNING CONCURRENTLY

I may be a scattered, distracted developer trying to do too damn much but that's my damn right. And I have 64 gigs of RAM so why shouldn't I be this way.

What I want is to be able to switch from app to app and make changes. This is important because some apps provide APIs which other apps rely on and having to figure out what thing is on what port, etc, is just plain distracting.

Foreman and Procfile.dev is a way around this but there's a major hitch in your getalong (as my Texas wife might say).

Here's a sample Procfile.dev for an app I'm building called Cartazzi which makes a developer's life easier:

web: bin/rails server -p 5000
css: yarn build:css --watch
js: yarn build --reload
# docker: docker-compose up

And here's a Profile.dev for another application called Poolwizard which helps you maintain your swimming pool:

#web: bin/rails server -p $PORT
web: bin/rails server -p 5700
css: yarn build:css --watch
js: yarn build --reload
# docker: docker-compose up
worker: bundle exec sidekiq

If you run Cartazzi and Poolwizard together then you're going to get crashes and here's the error:

❯ foreman start -f Procfile.dev
11:53 web.1    | started with pid 72877
11:53 css.1    | started with pid 72878
11:53 js.1     | started with pid 72879
11:53 worker.1 | started with pid 72881
11:53 js.1     | yarn run v1.22.5
11:53 css.1    | yarn run v1.22.5
11:53 css.1    | $ tailwindcss --postcss -i ./app/assets/stylesheets/application.tailwind.css -o ./app/assets/builds/application.css --watch
11:53 js.1     | $ node esbuild.config.js --reload
11:54 js.1     | node:events:371
11:54 js.1     |       throw er; // Unhandled 'error' event
11:54 js.1     |       ^
11:54 js.1     |
11:54 js.1     | Error: listen EADDRINUSE: address already in use :::5200
11:54 js.1     |     at Server.setupListenHandle [as _listen2] (node:net:1306:16)
11:54 js.1     |     at listenInCluster (node:net:1354:12)
11:54 js.1     |     at Server.listen (node:net:1441:7)
11:54 js.1     |     at buildAndReload (/Users/sjohnson/Sync/coding/rails/poolwizard/esbuild.config.js:54:6)
11:54 js.1     |     at Object.<anonymous> (/Users/sjohnson/Sync/coding/rails/poolwizard/esbuild.config.js:100:3)
11:54 js.1     |     at Module._compile (node:internal/modules/cjs/loader:1109:14)
11:54 js.1     |     at Object.Module._extensions..js (node:internal/modules/cjs/loader:1138:10)
11:54 js.1     |     at Module.load (node:internal/modules/cjs/loader:989:32)
11:54 js.1     |     at Function.Module._load (node:internal/modules/cjs/loader:829:14)
11:54 js.1     |     at Function.executeUserEntryPoint [as runMain] (node:internal/modules/run_main:76:12)
11:54 js.1     | Emitted 'error' event on Server instance at:
11:54 js.1     |     at emitErrorNT (node:net:1333:8)
11:54 js.1     |     at processTicksAndRejections (node:internal/process/task_queues:83:21) {
11:54 js.1     |   code: 'EADDRINUSE',
11:54 js.1     |   errno: -48,
11:54 js.1     |   syscall: 'listen',
11:54 js.1     |   address: '::',
11:54 js.1     |   port: 5200
11:54 js.1     | }
11:54 js.1     | error Command failed with exit code 1.
11:54 js.1     | info Visit https://yarnpkg.com/en/docs/cli/run for documentation about this command.
11:54 js.1     | exited with code 1
11:54 system   | sending SIGTERM to all processes
11:54 worker.1 | terminated by SIGTERM
11:54 css.1    | exited with code 1
11:54 web.1    | terminated by SIGTERM

The first time I looked at this, I saw the port error and I assumed that the error was the rails server because I knew that was running on a specific port. The actual error though here is node – which ISN'T EVEN IN THE Procfile.dev because it is run behind the scenes by yarn.

Here's the fix:

#web: bin/rails server -p $PORT
web: bin/rails server -p 5700
css: yarn build:css --watch
js: PORT=5701 yarn build --reload
# docker: docker-compose up
worker: bundle exec sidekiq

To understand this we need to put on our Unix underpinning's hat and think about this. You have Yarn executing and running Node as a process. Node inherits the environment variables that Yarn had which means if we set the PORT variable for Yarn then Node will inherit it.

The convention I'm adopting is this:

In all my Procfile.dev's, yarn build will get the port number of the Rails server + 1.

Thank You

I didn't figure this out on my own, a fellow Rails guy on Twitter, Kyle Keesling told me that this should work. That was enough for me to know that it could be figured out and to force me back into diagnostic mode. Thanks Kyle! Also Kyle is building a course on Stripe and Rails which looks good. Check it out here.

Thank You

See Also