Bridging the Safety Gap from Scripts to Full Auto-Remediation
summary
Dropbox builds an alert → diagnose → authorize → remediate pipeline: diagnose plugins re-check alerts every five minutes and emit a concrete prescription (reboot, deallocate, reformat), a human must approve via Human Authorized Execution, then pre-remediate hooks run automated checks (rate limits, availability, data-center isolation) before a remediate plugin executes. Failures re-enter the pipeline with a remediation-attempt counter. Teams remove the human gate per specific remediation only after protections prove reliable.
AI-written summary · about
chapters (9)
- 0:00Start
- 0:12Auto Remediation at Dropbox
- 3:56What Is the Road Map with Your Tools
- 6:13Human Authorization
- 10:49Symptom-Based Alert
- 11:59Pre-Remediate Hooks
- 12:36A Pre-Remediate Hook for a Zookeeper Server
- 18:10Race Conditions
- 19:38Rate Limits
related talks
20mSoftware and Hardware Remediations
32mAutomating Security Compliance at Scale
28mAuto-remediation at scale using watchers
29mComposing the Response: Building an Incident Pipeline from Scratch
43mBuilding a Secure DevOps Pipeline
54mBridging the Gap CICD Mind the Gap
45mSelf-Healing Pipelines? Resilient Deployments with Hardened Containers
34mI Break Things, AI Fixes Them: Building a Self-Healing CI/CD Pipeline