Go Away Or I Will Replace You With A Very Little Shell Script
summary
The talk examines how development and operations roles converge while differing success metrics still cause conflict, and describes using Infrastructure as Code (Puppet/Chef/Ansible/Salt), containers/Docker, monitoring (Monasca, Influx), feature flags, and separate deploy vs activation to manage that. It gives operational practices and metrics from Booking: ~10 deploys/day, capacity tests that cause a few 500 errors, budgeting downtime, blameless postmortems, and policies like “if you break it, can you fix it” to improve survivability and fault isolation.
AI-written summary · about
chapters (28)
- 0:00Introduction
- 1:14IT architecture evolution
- 4:24Scale-up vs scale-out
- 5:49Automation in practice
- 8:35DevOps as bottom-up
- 11:50Dev and ops collaboration
- 13:18Code deployment in production
- 16:29Dev vs ops metrics
- 18:39Booking.com rules
- 21:27Kernel project culture
- 22:32Crisis management at Web.de
- 24:19Rollouts and feature flags
- 26:51Downtime budget and errors
- 28:13Making failures survivable
- 30:21Production tests and capacity
- 32:09Testing system resilience
- 34:15DevOps conflicts persist
- 36:24Container complexity
- 38:32System layer complexity
- 40:39Sysadmins and SREs role
- 42:49DBAs and developer collab
- 45:20Capacity planning
- 47:05Migration to automation
- 49:53Management support needed
- 51:41Metrics across domains
- 52:49Testing in production
- 54:16ITIL vs DevOps
- 56:46Summary
related talks
34mDevops or: How I Learned to Stop Worrying and Love the Ops Team
33mOperations: a developer's guide
52mDevops @Scale
53mChange your tools, change your outcome
36mDevOps in Outsource… Really!
53mGround Truths of a Rugged DevOps Practitioner
40mDistilled Abnormal DevOps Iterations
49mDevOps Ten Years After: Review of a Failure with John Allspaw and Paul Hammond