.. in production

JSFoo '16

Bangalore

.. here to talk about the things we didn't give much fork about!

Logging, monitoring & alerts
Configurations
Profiling and control port
Clustering
Sticky sessions
Conventions

I'm Jai

Engineering @ DoSelect

@dolftax

Logging

What to log?

log, info, warn, error, fatal, .. is an overkill

What to log?

log, info, warn, error, fatal, .. is an overkill
Every log write affects performance

What to log?

log, info, warn, error, fatal, .. is an overkill
Every log write affects performance
`try { } catch (e) { }` does exist

Takeaway

Resist the tendency to log everything

Takeaway

Resist the tendency to log everything
The more you log, the less you can find

Takeaway

Resist the tendency to log everything
The more you log, the less you can find
`log`, `error`. Period.

Monitoring - The Obvious

IO? Why?

Because, libuv

Configuration

try {
  global.config = require('./dev-config')
} catch (err) {
  global.config = require('./config')
}

Key takeaway: Leave the configuration to deployer playbooks

Containers?

Take a sip of BudLight and ask yourself again

Mem leaks

Memory

Time

Heapdump helps!

// To generate heap snapshot
require('heapdump')

// Run garbage collector
process.on('SIGUSR2', function () {
  gc()
})

$ kill -USR2 <pid>

Use all the cores

Scale vertically first

Use all the cores

Scale vertically first
require('cluster') // && fork workers

.. cluster - bare minimum

if (cluster.isMaster) {
  // Fork workers.
  for (var i = 0; i < numCPUs; i++) {
    cluster.fork()
  }

  cluster.on('exit', (worker, code, signal) => {
    console.log(`worker ${worker.process.pid} died`)
  })
} else {
  // Listen
}

`cluster` ain't enough

Websockets?

Problem:

Websocket handshake might arrive at different workers

Solution:

Maintain sticky session when balancing requests

Sticky session

Conventions

doselect/bender/git.md
doselect/bender/javascript.md
Do *not* fight over semicolons

Questions?

Thanks much!

@dolftax