Telemetry

In this talk, we'll look at Telemetry, a new project for metrics and monitoring in the Elixir ecosystem. Telemetry provides libraries for gathering monitoring data from libraries and applications. It's already integrated with Ecto and is coming soon to Phoenix. We'll go through the design and thought process behind it, discuss the rationale for the not-so-obvious decisions, and most importantly, see the bright future of using a unified monitoring solution throughout the community.

Transcript

1. @arkgil @_arkgil TELEMETRY
2. @arkgil @_arkgil
3. @arkgil @_arkgil
4. @arkgil @_arkgil
5. @arkgil @_arkgil
6. @arkgil @_arkgil
7. @arkgil @_arkgil MONITORING
8. @arkgil @_arkgil
9. @arkgil @_arkgil
10. @arkgil @_arkgil Goals
11. @arkgil @_arkgil
12. @arkgil @_arkgil
13. @arkgil @_arkgil
14. @arkgil @_arkgil
15. @arkgil @_arkgil
16. @arkgil @_arkgil
17. @arkgil @_arkgil
18. @arkgil @_arkgil def handle_request(host, path, method, headers, body) do params = decode(body) case MyApp.business_logic(params) do {:ok, data} -> body = encode(data) send_resp(200, body) :error -> body = %{error: "Oops"} send_resp(400, body) end end
19. @arkgil @_arkgil def handle_request(host, method, body) do update_counter( “my_app.http.requests", host: host, method: method ) params = decode(body)
20. @arkgil @_arkgil case MyApp.business_logic(params) do {:ok, data} -> update_counter( “my_app.http.responses”, host: host, method: method, status: 200 )
21. @arkgil @_arkgil :error -> update_counter( “my_app.http.responses", host: host, method: method, status: 400 ) end
22. @arkgil @_arkgil def business_logic(params) do update_counter(“my_app.business_logic”, ...) with :ok <- authorize(params), {:ok, data} <- query_database(params), :ok <- push_event(data) do {:ok, data} end end
23. @arkgil @_arkgil def authorize(params) do update_counter("my_app.auth.requests", ...) ... end def query_database(params) do update_counter("my_app.db.queries", ...) ... end def push_event(params) do update_counter("my_app.broker.events_pushed", ...) ... end
24. @arkgil @_arkgil record_value("my_app.memory.total")
25. @arkgil @_arkgil
26. @arkgil @_arkgil
27. @arkgil @_arkgil def handle_request(host, method, body) do params = decode(body) update_counter(“my_app.http.requests”, ...) update_sum(“my_app.http.payload_size", ...) case MyApp.business_logic(params) do {:ok, data} -> update_counter(“my_app.http.responses”, ...) 
 update_statistics(“my_app.http.response_time”, ...) body = encode(data) send_resp(200, body) :error -> update_counter(“my_app.http.responses", ...) 
 update_statistics(“my_app.http.response_time”, ...) body = %{error: "Oops"} send_resp(400, body) end end
28. @arkgil @_arkgil The monitoring and reporting systems should be like an exoskeleton built around your system, not woven into it. - Michael Nygard, Release It!, 2nd ed.
29. EVENTS
30. @arkgil @_arkgil def handle_http_start( event, measurements, metadata, config ) do update_counter("my_app.http.requests", ...) update_sum( "my_app.http.payload_size", 
 measurements.payload_size, … ) end
31. @arkgil @_arkgil :telemetry
32. @arkgil @_arkgil :telemetry.attach( "my-handler", [:my_app, :http, :request, :start], &handle_http_start/4, nil )
33. @arkgil @_arkgil :telemetry.execute( [:my_app, :http, :request, :start], %{payload_size: payload_size}, %{host: host, method: method} )
34. @arkgil @_arkgil :telemetry.execute( [:my_app, :http, :request, :start], %{payload_size: payload_size}, %{host: host, method: method} )
35. @arkgil @_arkgil :telemetry.execute( [:my_app, :http, :request, :start], %{payload_size: payload_size}, %{host: host, method: method} )
36. @arkgil @_arkgil :telemetry.execute( [:my_app, :http, :request, :start], %{payload_size: payload_size}, %{host: host, method: method} )
37. @arkgil @_arkgil def handle_request(host, method, body) do params = decode(body) :telemetry.execute([:my_app, :http, :request, :start], ...) case MyApp.business_logic(params) do {:ok, data} -> body = encode(data) :telemetry.execute([:my_app, :http, :request, :stop], ...) send_resp(200, body) :error -> :telemetry.execute([:my_app, :http, :request, :stop], ...) body = %{error: "Oops"} send_resp(400, body) end end
38. @arkgil @_arkgil :telemetry.attach(“http-requests",[:plug, :request], ...) :telemetry.attach("db-queries", [:repo, :query], ...) :telemetry.attach("controllers", [:controller, :call], ...)
39. @arkgil @_arkgil
40. @arkgil @_arkgil
41. @arkgil @_arkgil
42. @arkgil @_arkgil Telemetry.Poller
43. @arkgil @_arkgil
44. @arkgil @_arkgil
45. @arkgil @_arkgil
46. @arkgil @_arkgil
47. @arkgil @_arkgil Telemetry.Metrics
48. @arkgil @_arkgil last_value("vm.memory.total")
49. @arkgil @_arkgil last_value("vm.memory.total") [:vm, :memory]
50. @arkgil @_arkgil last_value("vm.memory.total") [:vm, :memory] :total
51. @arkgil @_arkgil sum("http.request.payload_size") counter(“db.query.count") distribution("http.request.duration")
52. @arkgil @_arkgil last_value("vm.memory.total") sum("http.request.payload_size") counter("db.query.count") distribution("http.request.duration")
53. @arkgil @_arkgil last_value("vm.memory.total") sum("http.request.payload_size") counter("db.query.count") distribution("http.request.duration")
54. @arkgil @_arkgil last_value("vm.memory.total") sum("http.request.payload_size") counter("db.query.count") distribution("http.request.duration")
56. FUTURE
57. @arkgil @_arkgil
58. @arkgil @_arkgil
60. @arkgil @_arkgil
61. TELEMETRY
62. ..and metrics for all
63. @arkgil @_arkgil Links! - https://github.com/beam-telemetry/telemetry - https://github.com/beam-telemetry/telemetry_poller - https://github.com/beam-telemetry/telemetry_metrics - https://github.com/arkgil/telemetry_metrics_statsd - https://github.com/arkgil/rumbl-monitored