Welcome

This repository contains a colletion of challenges as part of the node performance workshop.

You can find a handy cheat-sheet below.

CHEATSHEET

Monitor Memory usage

$ ps -o pid,pmem,vsz,rss,comm -p <PID>

Alternatives: top, htop, free, vmstat

Self monitoring:

$ node
> process.memoryUsage()
{ rss: 13037568,
  heapTotal: 6147328,
    heapUsed: 2790824 }

Motitor CPU usage

$ ps -o pid,pcpu,cpu,comm -p <PID>

Alternatives: top, htop

Show CPU utilization over a 5 second period

$ ps -o pid,pcpu,cpu,comm -p <PID>

Show open TCP ports

lsof -nP -iTCP -sTCP:LISTEN

or

netstat -lntp

Show open file descriptors

Every file, port, unix pipe, or any kind of stream (unix stream) is represented by a file descriptor.

lsof -nP -p <PID>

Monitor network bandwith

iftop, iptraf, or nethogs

Monitor GC activity

node --trace_gc <program>

Monitor event loop lag

https://www.npmjs.com/package/toobusy-js

Write a heapdump every time RSS increases by 10 MB

var heapdump = require('heapdump');
var oneMb = Math.pow(2, 20);
var threshold = 10 * oneMb;
var last = get();
setInterval(check, 1000);
function get() { return process.memoryUsage().rss; }
function check() {
  var now = get();
  if (now - last < threshold)
    return;

  heapdump.writeSnapshot();
  console.log('Memory increase from %d MB to %d MB. Wrote dump',
              Math.floor(last / oneMb),
              Math.floor(now / oneMb));
  last = now;
}

Generate flamegraphs

Run your node program with:

node --perf_basic_prof myprogram.js

Run perf to trace the node process for 10 seconds:

sudo perf record -q -F 999 -p $(pgrep node) -g -- sleep 10

Make sure the generated map file is owned by the same user that runs perf:

sudo chown root /tmp/perf-*.map

Output the recorded activity into a file:

sudo perf script > /tmp/script

Use stackvis to generate a flame-graph from the perf script.

stackvis perf < /tmp/script > /tmp/flamegraph.html