Skip to main content
Version: Next

Migrate an Express Application to Watt

This guide takes an existing Express application and moves it into Watt in stages. Each stage is useful on its own and leaves you with a working application, so you can stop at whichever one gives you what you need.

StageWhat you changeWhat you get
1. WrapNothingWatt logging, metrics, management API, wattpm commands
2. Hand over the serverRemove app.listen, export buildMesh addressing, coordinated startup and shutdown
3. SplitMove code into multiple applicationsIsolation, per-application scaling

Nothing about your routes, middleware, or error handlers changes at any stage. Express stays Express.

The application we are starting from​

A typical Express service — routes, a listen call at the bottom, nothing unusual:

// index.js
import express from 'express'

const app = express()

app.get('/health', (req, res) => {
  res.json({ status: 'ok' })
})

app.get('/users/:id', (req, res) => {
  res.json({ id: req.params.id, name: 'Ada Lovelace' })
})

const port = process.env.PORT ?? 3000
app.listen(port, () => {
  console.log(`listening on ${port}`)
})

Stage 1: Wrap it, change nothing​

Run the Watt creator from the root of your existing project:

npm create wattpm

It detects that the directory already contains a Node.js application and offers to wrap it:

Hello YOURNAME, welcome to Watt!
? This folder seems to already contain a Node.js application. Do you want to wrap into Watt? yes
- Installing @platformatic/runtime using npm ...
INFO: .env written!
INFO: .env.sample written!
INFO: package.json written!
INFO: watt.json written!
INFO: Installing dependencies for the project using npm ...
INFO: You are all set! Run `npm start` to start your project.

This writes a watt.json that treats your project as a single @platformatic/node application:

{
  "$schema": "https://schemas.platformatic.dev/@platformatic/node/3.0.0.json",
  "runtime": {
    "logger": {
      "level": "{PLT_SERVER_LOGGER_LEVEL}"
    },
    "managementApi": "{PLT_MANAGEMENT_API}"
  }
}

Start it:

npx wattpm dev

Your app.listen() call still runs. Watt notices the server your application started and adopts it:

Starting the worker 0 of the application "legacy-express"...
listening on 3000
Started the worker 0 of the application "legacy-express"...
Platformatic is now listening at http://127.0.0.1:3000

Your routes work exactly as before, and your console.log output is captured into the structured log stream. You now have Watt's logging, metrics, health probes, and management API without having touched a line of application code.

note

The wrap step does not rewrite your package.json start script. If it was node index.js, it still is. Change it to wattpm start (and add "dev": "wattpm dev") when you want npm start to go through Watt.

Stage 2: Let Watt own the server​

Stage 1 works, but your application is still opening its own TCP socket. Exporting a build function instead hands the server to Watt, which is what unlocks the rest of the platform.

Delete the listen call and export the app:

// index.js
import express from 'express'

export function build () {
  const app = express()

  app.get('/health', (req, res) => {
    res.json({ status: 'ok' })
  })

  app.get('/users/:id', (req, res) => {
    res.json({ id: req.params.id, name: 'Ada Lovelace' })
  })

  return app
}

Return the Express app itself — not an http.Server. Watt recognises it and wires it up.

What this buys you:

  • No TCP for internal traffic. Only the entrypoint gets a real socket. Everything else is reached over the mesh, which is a message between threads rather than a network round trip.
  • Coordinated lifecycle. Watt starts your application in dependency order and shuts it down gracefully rather than pulling the socket out from under in-flight requests.
  • Restarts and health checks. See Application Lifecycle.

The name can be build or create; both are recognised.

Stage 3: Use the runtime logger​

Your application's own logger still works, but routing it through Watt means every application's logs land in one stream, already tagged with the application id and worker index:

import { getLogger } from '@platformatic/globals'
import express from 'express'

export function build () {
  const app = express()
  const logger = getLogger()

  app.get('/users/:id', (req, res) => {
    logger.info({ id: req.params.id }, 'fetching user')
    res.json({ id: req.params.id, name: 'Ada Lovelace' })
  })

  return app
}

Which produces:

{"level":30,"name":"api","worker":0,"id":"42","msg":"fetching user"}

getLogger() returns a Pino logger, so if you were already using Pino the call signatures are unchanged. See Logging for the full configuration surface.

Stage 4: Split into multiple applications​

This is the stage that pays for the previous ones. Move each Express application into its own directory under web/:

my-project/
├── package.json          # "workspaces": ["web/*"]
├── watt.json             # runtime configuration
└── web/
    ├── api/              # the Express application
    │   ├── package.json
    │   ├── watt.json
    │   └── index.js
    └── notifier/
        ├── package.json
        ├── watt.json
        └── index.js

The runtime configuration picks up everything under web/ and names one application as the entrypoint:

{
  "$schema": "https://schemas.platformatic.dev/wattpm/3.0.0.json",
  "entrypoint": "api",
  "autoload": { "path": "web" },
  "server": { "hostname": "127.0.0.1", "port": 3042 },
  "logger": { "level": "info" }
}

Each application gets a minimal watt.json naming its capability:

{
  "$schema": "https://schemas.platformatic.dev/@platformatic/node/3.0.0.json"
}

Applications call each other over the mesh, using the directory name as the hostname:

app.get('/users/:id', async (req, res) => {
  const response = await fetch(`http://notifier.plt.local/greeting/${req.params.id}`)
  const { greeting } = await response.json()

  res.json({ id: req.params.id, name: 'Ada Lovelace', greeting })
})

No port, no service registry, no environment variable holding an address. The application id is the address.

This is ordinary fetch against an ordinary URL, which is the point: if you later extract notifier into its own deployment, this call site changes only in hostname. See The Modular Monolith.

Putting an application behind a gateway​

When an Express application is not the entrypoint, it usually sits behind a gateway that exposes it under a prefix.

Declare the prefix on the application:

{
  "$schema": "https://schemas.platformatic.dev/@platformatic/node/3.0.0.json",
  "application": {
    "basePath": "/orders"
  }
}

and let the gateway pick it up:

{
  "$schema": "https://schemas.platformatic.dev/@platformatic/gateway/3.0.0.json",
  "gateway": {
    "applications": [{ "id": "orders" }],
    "refreshTimeout": 1000
  }
}
The prefix is stripped before your application sees it

By default the gateway removes the prefix on the way through. A request for /orders/health arrives at your Express app as /health.

This is what you want when migrating: your existing routes keep working unchanged. Do not add the prefix to your route definitions — if you register /orders/health, the incoming /health will not match it and you will get a 404 that looks like a routing bug.

If your application genuinely needs the full path — because it builds absolute URLs, or generates links that must include the prefix — set absoluteUrl and prefix your routes explicitly:

{
  "$schema": "https://schemas.platformatic.dev/@platformatic/node/3.0.0.json",
  "application": { "basePath": "/orders" },
  "node": { "absoluteUrl": true }
}
import { cleanBasePath } from '@platformatic/basic'
import { getBasePath } from '@platformatic/globals'

const prefix = getBasePath({ throwOnMissing: false }) ?? ''

app.get(cleanBasePath(`${prefix}/health`), (req, res) => {
  res.json({ status: 'ok' })
})

Pick one mode or the other. Mixing them — absoluteUrl off with prefixed routes, or on with unprefixed routes — produces 404s in both directions.

Production​

npm run build   # wattpm build
npm run start   # wattpm start

wattpm build runs each application's build step, if it has one. For a plain Express application there is usually nothing to build, but running it keeps the command identical across a project that also contains frontends.

Things that catch people out​

  • process.env.PORT no longer decides anything. Watt assigns ports. The entrypoint's public port comes from server.port in the runtime configuration; non-entrypoint applications normally have no port at all.
  • In-memory state is per worker. If you give an application multiple workers, each has its own heap. Sessions, caches, and rate-limit counters need a shared store — exactly as they would across containers. See The Multithread Model.
  • process.exit() takes down the whole runtime. Threads share a process. Throw or return an error instead.
  • process.cwd() is the runtime root, not your application directory. Resolve paths relative to import.meta.dirname.
  • Only the entrypoint is reachable from outside. This is a feature — a non-entrypoint application does not open a listening socket — but it does mean you cannot curl it directly. Use wattpm inject or route through the gateway.