際際滷

際際滷Share a Scribd company logo
Command & Conquer
!
息 Tadej Murovec, HouseTrip, 2015 1
! Hello
Twitter: @tadejm
GitHub: github.com/tadejm
Work at HouseTrip.
息 Tadej Murovec, HouseTrip, 2015 2
!
息 Tadej Murovec, HouseTrip, 2015 3
! The debt
184 alerts per week
=> 1.095 alert per hour
息 Tadej Murovec, HouseTrip, 2015 4
息 Tadej Murovec, HouseTrip, 2015 5
! Rethink how to do alerting
 Alerts should be urgent, important, actionable and real.
 Over-monitoring is a harder problem to solve than under-
monitoring.
 Symptoms are a better way to capture more problems
more comprehensively and robustly with less effort.1
1
My Philosophy on Alerting  Rob Ewaschuk
息 Tadej Murovec, HouseTrip, 2015 6
! Meet the Happy path
* Guests can search for properties
* Guests can browse properties
* Users can login or register
* Guests can send an enquiry or book a property
* Guests can pay
* Hosts can accept booking
息 Tadej Murovec, HouseTrip, 2015 7
The end user does not care about MySQL server being
unreachable, but she does care about not being able to
view a property.
息 Tadej Murovec, HouseTrip, 2015 8
Tools & services
息 Tadej Murovec, HouseTrip, 2015 9
息 Tadej Murovec, HouseTrip, 2015 10
New Relic
IS AWESOME.
息 Tadej Murovec, HouseTrip, 2015 11
New Relic
 Application server response time
 Key transactions tracking with Apdex T
 Uptime monitoring
 Application segmentation into policy groups
息 Tadej Murovec, HouseTrip, 2015 12
息 Tadej Murovec, HouseTrip, 2015 13
Datadog
IS AWESOME.
息 Tadej Murovec, HouseTrip, 2015 14
Datadog
 Monitoring scheduled errands
 Background queues and services health check
 Custom metrics that make sense for the business
息 Tadej Murovec, HouseTrip, 2015 15
息 Tadej Murovec, HouseTrip, 2015 16
Slack
IS AWESOME.
息 Tadej Murovec, HouseTrip, 2015 17
Slack
 Chronological tracking, search
 Slack channels as urgency segmentation
#alerts, #notifications
息 Tadej Murovec, HouseTrip, 2015 18
息 Tadej Murovec, HouseTrip, 2015 19
PagerDuty
IS OK.
And wakes you up at night. Sometimes.
息 Tadej Murovec, HouseTrip, 2015 20
PagerDuty
 Propagates alerts to duty engineers
 Manage duty rotation schedule
 Keep notes about alerts
息 Tadej Murovec, HouseTrip, 2015 21
! Progress and results
息 Tadej Murovec, HouseTrip, 2015 22
! 4 rules for efficient alerting
 Alert on symptoms not causes
" Get your teammates to pair review the alerts
# Prefer under monitoring to over monitoring
$ Use noti鍖cations to prevent alerts
息 Tadej Murovec, HouseTrip, 2015 23
! Be a good citizen
息 Tadej Murovec, HouseTrip, 2015 24
息 Tadej Murovec, HouseTrip, 2015 25
That's all folks!
Thanks! !
息 Tadej Murovec, HouseTrip, 2015 26

More Related Content

Command & conquer: Red alert

  • 1. Command & Conquer ! 息 Tadej Murovec, HouseTrip, 2015 1
  • 2. ! Hello Twitter: @tadejm GitHub: github.com/tadejm Work at HouseTrip. 息 Tadej Murovec, HouseTrip, 2015 2
  • 3. ! 息 Tadej Murovec, HouseTrip, 2015 3
  • 4. ! The debt 184 alerts per week => 1.095 alert per hour 息 Tadej Murovec, HouseTrip, 2015 4
  • 5. 息 Tadej Murovec, HouseTrip, 2015 5
  • 6. ! Rethink how to do alerting Alerts should be urgent, important, actionable and real. Over-monitoring is a harder problem to solve than under- monitoring. Symptoms are a better way to capture more problems more comprehensively and robustly with less effort.1 1 My Philosophy on Alerting Rob Ewaschuk 息 Tadej Murovec, HouseTrip, 2015 6
  • 7. ! Meet the Happy path * Guests can search for properties * Guests can browse properties * Users can login or register * Guests can send an enquiry or book a property * Guests can pay * Hosts can accept booking 息 Tadej Murovec, HouseTrip, 2015 7
  • 8. The end user does not care about MySQL server being unreachable, but she does care about not being able to view a property. 息 Tadej Murovec, HouseTrip, 2015 8
  • 9. Tools & services 息 Tadej Murovec, HouseTrip, 2015 9
  • 10. 息 Tadej Murovec, HouseTrip, 2015 10
  • 11. New Relic IS AWESOME. 息 Tadej Murovec, HouseTrip, 2015 11
  • 12. New Relic Application server response time Key transactions tracking with Apdex T Uptime monitoring Application segmentation into policy groups 息 Tadej Murovec, HouseTrip, 2015 12
  • 13. 息 Tadej Murovec, HouseTrip, 2015 13
  • 14. Datadog IS AWESOME. 息 Tadej Murovec, HouseTrip, 2015 14
  • 15. Datadog Monitoring scheduled errands Background queues and services health check Custom metrics that make sense for the business 息 Tadej Murovec, HouseTrip, 2015 15
  • 16. 息 Tadej Murovec, HouseTrip, 2015 16
  • 17. Slack IS AWESOME. 息 Tadej Murovec, HouseTrip, 2015 17
  • 18. Slack Chronological tracking, search Slack channels as urgency segmentation #alerts, #notifications 息 Tadej Murovec, HouseTrip, 2015 18
  • 19. 息 Tadej Murovec, HouseTrip, 2015 19
  • 20. PagerDuty IS OK. And wakes you up at night. Sometimes. 息 Tadej Murovec, HouseTrip, 2015 20
  • 21. PagerDuty Propagates alerts to duty engineers Manage duty rotation schedule Keep notes about alerts 息 Tadej Murovec, HouseTrip, 2015 21
  • 22. ! Progress and results 息 Tadej Murovec, HouseTrip, 2015 22
  • 23. ! 4 rules for efficient alerting Alert on symptoms not causes " Get your teammates to pair review the alerts # Prefer under monitoring to over monitoring $ Use noti鍖cations to prevent alerts 息 Tadej Murovec, HouseTrip, 2015 23
  • 24. ! Be a good citizen 息 Tadej Murovec, HouseTrip, 2015 24
  • 25. 息 Tadej Murovec, HouseTrip, 2015 25
  • 26. That's all folks! Thanks! ! 息 Tadej Murovec, HouseTrip, 2015 26