Elug SP 2016 - Measuring your Elixir Application

In this talk I show how you can use Exometer, InfluxDB and Grafana to collect & observer application metrics easily and extensibly.

Transcript

1. MEASURING YOUR ELIXIR APPLICATION Renan Ranelli
2. … [OR]: HOW TO USE NUMBERS TO MAKE BETTER DECISIONS Renan Ranelli
3. Milhouse (Renan Ranelli)
4. Software Engineer @ Milhouse (Renan Ranelli)
5. (former) Software Engineer @ Milhouse (Renan Ranelli)
6. Milhouse (Renan Ranelli)
7. AGENDA • Why should I care about metrics? • General principles when collecting metrics • How we measure our things @ Xerpa • Demo demo demo
8. DISCLAIMER • The main inspiration for this talk is Coda Hale's “Metrics Metrics” talk: • https://www.youtube.com/watch?v=czes-oa0yik
9. WHY SHOULD I CARE ABOUT METRICS?
10. WHY? • We write code.
11. WHY? • We write code. • Code generates business value.
12. WHY? • We write code. • Code generates business value. – When we RUN it is running. – Not when we write it.
13. WHY? • In order to generate more business value, we need to make better decisions about our code.
14. WHY? • In order to generate more business value, we need to make better decisions about our code. • In order to make better decisions about our code (about anything really) we need to measure it.
15. THE MAP IS NOT THE TERRITORY
16. WHY? • We have a mental model of what our code does.
17. WHY? • We have a mental model of what our code does. • We are often wrong
18. WHY? • We have a mental model of what our code does. • We are often wrong • There is no physical counterpart for the things we work with. We spend too much times inside our heads.
19. WHICH IS FASTER?
20. WE DON'T KNOW
21. WHY?
22. WHY?
23. WHY? • If we we're so wrong in such a trivial example, imagine a typical web-app scenario:
24. WHY? • If we we're so wrong in such a trivial example, imagine a typical web-app scenario: – Action routing – Database queries – Template rendering – Session cache
25. WHY? • If we we're so wrong in such a trivial example, imagine a typical web-app scenario: – Action routing (10ms) – Database queries (31ms) – Template rendering (22ms) – Session cache (327ms)
26. WE DON'T KNOW
27. WE DON'T KNOW UNTIL WE SEE OUR CODE FOR WHAT IT **ACTUALLY** DOES (SCIENCE!!)
28. METRICS METRICS METRICS
29. TYPES OF MEASUREMENTS – Counters – Gauges – Histograms – Timers (See coda-hales “metrics, metrics” talk!)
30. TYPES OF MEASUREMENTS
31. HOW WE MEASURE THINGS @ XERPA
32. •
33. •
34. •
35. • ESTAMOS CONTRATANDO
36. • ESTAMOS CONTRATANDO (principalmente se vc manjar frontend)
37. • ESTAMOS CONTRATANDO (principalmente se vc manjar frontend, mas também rola backend)
38. •
39. • • • SQL-like query language • HTTP interface • Continuous Queries • Seamless integration with Grafana • (has an exometer reporter)
40. • • <<<<<<<< LIVE DEMO >>>>>>>>> • https://github.com/rranelli/elugsp_milmetrics
41. • mix.exs
42. • config.exs
43. • plug/metrics.exs
44. • Start `grafana` and `influxdb` with `make start` • Configure a new data source for influxdb 0.9x, user: root, pass: root, host: http://localhost:8086 • Profit
45. OBRIGADO !
46. @renanranelli /rranelli Renan Ranelli (Milhouse) milhouseonsofware.com