One of my many failings as a developer is the cadence to onboard a new tool. pm2 came to me as an alternative to supervisord without the live console updates. I knew of its strengths, but failed to utilize it.

Reviewing my Strider setup which expects config via environment variables, I have new motivation to ramp up on pm2.

Key likes about pm2:

Manages node instances to optimize CPU core availability.
Manages service start-up/ shutdown with environment config.
Manages code deployment from development to production.

Learnings on the above.

How does a multi-instance service work?

App performance can be boosted by configuring pm2 to launch multiple instances, limited by available cores (cool right?). The following code responds once to a GET call and ends. Save it in a file (eg. testinstance.js).

const restify = require('restify')

const server = restify.createServer({
    name: 'testinstance',
    version: '1.0'
})

server.get('/', (req, res, next) => {
    res.send({
        InstanceId: process.env.NODE_APP_INSTANCE,
        PmId: process.env.pm_id
    }, {
        Connection: 'close'
    })
    // console.log(server.getDebugInfo())
    server.close(() => {
        console.log('Service stopped')
    })
    next();
})

server.listen(8000, () => {
    console.log('%s listening at %s', server.name, server.url)
})

Run it in cluster mode inside pm2:

pm2 start testinstance.js -i max

Notice there is no additional code required to make the HTTP service multi-core aware. Hitting the url will get a response similar to the following:

{"InstanceId":"0","PmId":"3"}

Call the url and notice how the PmId changes each time until the service stops responding. That's when all the instances have stopped.

::: warning Question pm2 continues to report stopped instances as 'online'. Why will it not trigger a restart? Is it expecting a certain exit code? :::

How does a multi-instance service work?

References