Putting the Ops in GitOps - New Challenges in Managing Cloud Native Environments
summary
GitOps speeds deployments but introduces concrete operational costs: pipelines remain imperative and can add 40+ minute delays to emergency fixes, rollbacks rarely handle database migrations cleanly, and partial multi-service deploys leave systems in indeterminate states. The talk documents real mitigations — using an infrastructure lock during incidents, allowing controlled infrastructure drift for urgent remediation, preferring the original two-repo pattern rather than a single monorepo, and integrating reconcilers like Argo CD to automate resyncs.
AI-written summary · about
chapters (26)
- 0:00Intro
- 1:24Analogy of OneWay Road
- 4:40What is GitOps
- 6:00GitOps Magic Spiral
- 7:30GitOps Criticism
- 9:07Organization
- 9:44Written Rules
- 11:07Organizational Reasons
- 12:00Criticism
- 12:52Pipelines
- 13:33Microservice breakdown
- 13:54Deploy is not momentary
- 15:05What do you do about it
- 15:40Rollback
- 16:28Delayed Problem Detection
- 17:22The reality exists outside the repo
- 18:18Infrastructure Drift
- 20:00Infrastructure Lock
- 21:12GitOps as a Vision
- 21:27How GitOps makes life harder
- 22:13GitOps didnt solve the devops problem yet
- 24:00Lets become philosophical
- 24:10Jack of all trades
- 26:04Master of none
- 27:21Summary
- 27:54What is next
related talks
55mGitOps, Fostering Next-Generation Operations Teams
29mThe Evolution of GitOps in Platform Engineering
31mPanel Discussion: GitOps & Best Practices for Cloud Native CI/CD
29mManaging infrastructure with Git: First principles of GitOps
05mCloud Native CI/CD using 'GitOps'
30mOrganizing Teams for GitOps and Cloud Native Deployments
13mHow GitOps Should I Be? Lessons from a Kinda-GitOps Environment
40mHow GitOps Changes Everything in CI/CD