CODEWITHSUNDEEP
deploymentloopbackjsstrongloop
notbrain
notbrain

Reputation: 3396

strongloop slc deploy env var complications

I've been deploying a loopback app via a custom init.d/app.conf script, using slc run --detach --cluster "cpu", but want to move to using strong-pm, as recommended.

But I've come across some limitations and am looking for any guidance on how to replicate the setup with which I'm currently familiar.

Currently I set app-specific configuration inside server/config.local.js and server/datasources.local.js, most importantly the PORT at which the app should listen for connections on. This works perfectly using slc run for local development and remote deploying for staging, all I do is set different env vars for each distinct app:

datasources.local.js:

module.exports = {
  "mysqlDS": {
    name: "mysqlDS",
    connector: "mysql",
    host: process.env.PROTEUS_MYSQL_HOST,
    port: process.env.PROTEUS_MYSQL_PORT,
    database: process.env.PROTEUS_MYSQL_DB,
    username: process.env.PROTEUS_MYSQL_USER,
    password: process.env.PROTEUS_MYSQL_PW
  }
}

config.local.js:

module.exports = {
  port: process.env.PROTEUS_API_PORT
}

When I deploy using strong-pm, I am not able to control this port, and it always gets set to 3000+N, where N is just incremented based on the service ID assigned to the app when it's deployed.

So even when I deploy and then set env using

slc ctl -C http://localhost:8701 env-set proteus-demo PROTEUS_API_PORT=3033 PROTEUS_DB=demo APP_DOMAIN=demo.domain.com

I see that strong-pm completely ignores PROTEUS_API_PORT when it redeploys with the new env vars:

ENV has changed, restarting
Service "1" listening on 0.0.0.0:3001
Restarting next commit Runner: commit 1/deploy/default/demo-deploy
Start Runner: commit 1/deploy/default/demo-deploy
Request (status) of current Runner: child 20066 commit 1/deploy/default/demo-deploy
Request {"cmd":"status"} of Runner: child 20066 commit 1/deploy/default/demo-deploy

3001! Not 3033 like I want, and spec'd in config.local.js. Is there a way to control this explicitly? I do not want to need to run an slc inspection command to determine the port for my nginx upstream block each time I deploy an app. Would be awesome to be able to specify listen PORT by service name, too.

FWIW, this is on an aws instance that will host demo and staging apps pointing to separate DBs and on different PORTs.

Upvotes: 2

Views: 794

Answers (1)

Ryann Graham
Ryann Graham

Reputation: 8229

strong-pm only sets a PORT environment variable, which the app is responsible for honouring.

Based on loopback-boot/lib/executor:109, it appears that loopback actually prefers the PORT environment variable over the value in the config file. In that case it seems your best bet is to either:

  • pass a port in to app.listen() yourself
  • set one of the higher priority environment variables such as npm_config_port (which would normally be set via npm start --port 1234).

Upvotes: 5

Related Questions