Download as PDF, PPTX
More Related Content
Similar to Serverless computing henry been - logging instrumentation dashboards alerts
Serverless computing henry been - logging instrumentation dashboards alerts
- 1.
- 2. WONDERING WHO IS THATGUY? HENRY BEEN Independent Devops & Azure Architect E: henry@azurespecialist.nl T: @henry_been L: linkedin.com/in/henrybeen W: henrybeen.nl
- 3. Very boring part Prettyboring part Maybe boring part Amazing part Really amazing part Mindblowing part LOGGING, INSTRUMENTATION, DASHBOARDS & ALERTS
- 4. A bit ofmonitoring theory What to monitor Demo’s! LOGGING, INSTRUMENTATION, DASHBOARDS, ALERTS Continuous Deployment for Azure Serverless - @henry_been
- 5. MONITORING -or- KNOWING WHAT BEHAPPENING
- 6. Instrumentation Most often forgotten Concernedwith metrics • (Timestamp, value) tuples • Events & values • Dimensions • Used for Dashboards & Alerts Rolled up over time Logging Often combined with tracing Concerned with state changes • When • By who • What changed • Can be a source of metrics Retained for longer or indefinitly Tracing Used for troubleshooting What we developers call logging • When • Code location • Very, very verbose message • Can be a source of metrics Retained for a limited time MONITORING
- 7. VM / container Timeseries Database ApplicationInstance Agent SDK Logging Database Https endpoint Visualization Engine Your Dashboards Query Engine Your Alerts Streaming Data Engine
- 8. BUT HEY THATSTUFF IS OPERATIONS… WHY WOULD I CARE?
- 9.
- 10. • DevOps isthe union of people, process, and products to enable continuous delivery of value to your end users. “ ” DEVOPS
- 11. OKAY, OKAY ICARE… NOW, WHAT TO MONITOR?
- 12. OKAY, OKAY ICARE… NOW, WHAT NOT TO MONITOR?
- 13. WHAT NOT TOMONITOR (MAYBE) FOCUS ON AVERAGES Averages are lying cheats and hide real issues! • Use dimensions to find outliers • Use maximums or percentiles to identify spread FOCUS ON YOUR SYSTEMS Do you think your users care? At all? • Is the amount of disk IO really interesting? • Does page file usage predict conversion? To add some nuances: - Some system metrics are handy, just don’t start there maybe - Same goes for averages
- 14. 0 20 40 60 80 100 120 Database load(%) DO WEHAVE AN ISSUE? Proposed alert
- 15. ARE MY USERSHAPPY? 0 20 40 60 80 100 120 140 160 Average response time
- 16. 0 200 400 600 800 1000 1200 Average response timeand maximum response time Average response time Maximum response time ARE MY USERS HAPPY?
- 17. 0 100 200 300 400 500 600 700 800 900 Average response timeand response time per operation Average response time View item AddToBasket ARE MY USERS HAPPY?
- 18.
- 19. MEASURE THINGS THE USERCARES ABOUT AND THINK IT THROUGH!
- 20. Let’s do a QuickDemo! Demo
- 21. A VERY GOODBOOK ALSO QUITE THIN
- 22. FOR THOSE USINGAZURE HOW WOULD YOU DO IT?
- 23. AZURE SERVICES Azure Monitor ApplicationInsights Log Analytics
- 24.
- 25.
- 26. Platform Built into (almost)every offering Fixed metrics • Different for each offering • Automatically recorded Enabled by default and FREE! Custom Using Application Insights Custom metrics • Self-chosen name • Double values • Dimensions disabled by default Use the TelemetryClient class Application Using Application Insights Well known metrics • CPU percentage • Response times • Memory usage Install NuGet package, configure instrumation key and DONE METRIC - SOURCES
- 27. DEMO TIME! A TodoApplication… ALMOST
- 28. POST UrlShortner AddAliasProcessor SQLDatabase GET UrlShortner
- 29. VM / container Timeseries Database ApplicationInstance Agent SDK Logging Database Https endpoint Visualization Engine Your Dashboards Query Engine Your Alerts Streaming Data Engine
- 30. Two more things “YOUCAN USE THE DEMOS!” “AND YOU CAN USE THE DEMOS!” “EVERYBODY CAN USE THE DEMOS!” DASHBOARDS AND ALERTS AS CODE
- 31. DO TRY THISAT HOME! HENRY BEEN Independent Devops & Azure Architect E: henry@azurespecialist.nl T: @henry_been L: linkedin.com/in/henrybeen W: henrybeen.nl
- 32.