The Truth Is Out There... Monitoring Your Isomorphic Application

Going a full circle from web-masters to HTML developers and expert JavaScript engineers, more product teams are closing the loop by working towards full-stack JS teams capable of implementing whole (or a most of) features from scratch and even deploy it to customer facing servers. Stepping up the stack and doing more Node.js to power your back-ends, what could be more important than a good night sleep with proper monitoring set-up?

Robert will guide you through monitoring essentials to make sure your isomorphic JS application, or a set of microservices, are performing stable and power your business without hikups. Learn about tools, isomorphic logging solutions, performance and stability measurement approaches and stay confident in your uptimes.

Talk from http://css-minsk-js.by

Transcript

1. The Truth Is Out There...
   Robert Haritonov

   Tech Lead @
   Monitoring Your Isomorphic
   Application
   

   View Slide

2. Zoover.nl is a travel website
   
   3-4M
   Users a 

   month
   10+
   Years on the
   market
   

   View Slide

3. Rebuild everything from scratch
   
   
   

   View Slide

4. What is monitoring?
   • Measuring
   • Recording
   • Alerting
   • Visualising
   

   View Slide

5. Isomorphic == Fullstack
   

   View Slide

6. Full-stack culture
   
   

   View Slide

7. Monitoring:
   Why should you care?
   • Be confident in your product performance
   

   View Slide

8. "Our code generates business value
   when it runs, not when we write it"
   - Coda Hale,
   "Metrics, metrics, everywhere"
   

   View Slide

9. Monitoring:
   Why should you care?
   • Be confident in your product performance
   • Stay one step ahead, fix bugs before they affect UX
   • Good night sleep
   

   View Slide

10. What to monitor?
    • Business and tech performance
    • Error and Success
    • Throughput and availability metrics
    

    View Slide

11. Logging and Error Tracking
    
    
    

    View Slide

12. "Anything that can go wrong
    will go wrong"
    - Murphy's law
    

    View Slide

13. Error tracking services
    • Rollbar, Sentry, Bugsnag, New Relic...
    • Allow good discovery tools and alerting
    • Custom and auto-grouping
    • Source Maps, version tracking, and more...
    

    View Slide

14. Client-side logging
    • Propagate errors to tracking services (or own storage)
    • Log user actions for debugging and analytics
    • Provide useful context with errors (status code, description)
    • Keep the stack traces
    

    View Slide

15. "Undefined is not a function"
    - Your favourite browser
    

    View Slide

16. Bad
    

    View Slide

17. Good ✅
    

    View Slide

18. Server-side logging
    • Log in JSON with most context data you can get:
    • Code version, host, timestamp
    • Request parameters (logging middlewares)
    • Loggers like Bunyan support streams for log propagation
    

    View Slide

19. Streaming Logs
    

    View Slide

20. Isomorphic Logger
    

    View Slide

21. Isomorphic Logger
    

    View Slide

22. Isomorphic Logger
    

    View Slide

23. Isomorphic Logger
    

    View Slide

24. Isomorphic Logger
    

    View Slide

25. Notes from the field
    • International messages
    • Default Grouping could go wrong
    • Configure events throttling
    • Google bots from real devices will spawn uncommon errors
    

    View Slide

26. Configuration Tips
    

    View Slide

27. Configuration Tips
    

    View Slide

28. Configuration Tips
    

    View Slide

29. Throughput and Availability Metrics
    ⏱
    
    

    View Slide

30. "If it could affect your code’s 

    business value, add a metric."
    - Coda Hale,
    "Metrics, metrics, everywhere"
    

    View Slide

31. Tools We Use
    • Prometheus - metrics aggregator
    • Grafana - data visualizations
    • Pingdom - uptime monitoring
    

    View Slide

32. View Slide

33. Prometheus
    • Exposes `/metrics` endpoint
    • Supports numerous clients, including Node.js
    • Exports default metrics - CPU, Memory, Event Loop Lag
    • And custom - Counter, Gauge, Histogram
    

    View Slide

34. Default metrics
    

    View Slide

35. View Slide

36. Custom Metrics
    • Counter - for values only going up (req count, etc)
    • Gauge - instantaneous value of something (hardware metrics)
    • Histogram - the statistical distribution of values (req duration, etc)
    

    View Slide

37. Count Rendered Documents
    

    View Slide

38. Count Rendered Documents
    

    View Slide

39. Count Render Times
    

    View Slide

40. Count Render Times
    

    View Slide

41. More Prometheus Exporters
    • Redis, Memcached
    • Elasticsearch, PostgreSQL
    • Thumbor (image processing)
    • Google Analytics Reporting - infinityworks/gar-exporter
    

    View Slide

42. Alerting
    
    
    

    View Slide

43. Alerting
    • Pagerduty - alerts aggregations
    • Duty calendar
    

    View Slide

44. Duty Calendar
    

    View Slide

45. Alerting
    • Pagerduty - alerts aggregations
    • Duty calendar
    • Integration with Rollbar, Pingdom and other
    

    View Slide

46. Grafana Alerting
    

    View Slide

47. "Anything that can go wrong
    will go wrong"
    - Murphy's law
    

    View Slide

48. View Slide

49. Thank you!
    Robert Haritonov

    Tech Lead @
    

    View Slide