Video - Key to SRE

Excerpt (WIP)

• define written SLAs, SLOs for every service (and measure it!)
• Use error budgets as your launch criterea
• teams are self-policing - because they are not monolithic
• put a 50% cap on Ops work (tickets, maintanence, …)
  • usually ~30%
• on call also for devs
• SRE Portability (https://youtu.be/n4Wf14e2jxQ?t=2387)
  • no requirement to stick with any project
  • no requirement to stick with SRE
  • build it and they will come
  • bust it, and they will leave
• availability <100%! -> there will be outages!
  • minimize damage
  • practice, practice, practice (wheel of misfortune)
  • do post mortems

Prevent recurrenence of outages

1. Handle the event
2. Write post-mortems
3. Reset

Consequence

• max of 2 events per oncall shift
• minimum group size 8(people)x1(locations) or 6x2

Post Mortem philosophy

• PMs are blameless
• Assure people are intelligent, well-intentioned (“it’s not the person - it’s the environment”)
• Focus on process & tech
• create a timeline
• get all the facts
• Create bugs for all followup work

…

Links

• https://www.youtube.com/watch?v=n4Wf14e2jxQ