Live Q&A Session #2
summary
A live panel explains concrete reliability practices: a regional config validation system (RCV) that uses real user traffic and canaries to catch site-scale outages, tuned metrics that currently use static thresholds or comparison checks, and an empirical 10-minute window chosen to gather enough data without blocking developers. The pipeline includes unit tests for the consuming layer, randomized regional rollouts to auto-correlate bad metrics to releases, and just-in-time migrations with background disk migration to achieve near zero downtime while rollbacks remain manual and contextual.
AI-written summary · about
chapters (17)
- 0:00Introduction
- 0:47Biggest changes after 2021 outages
- 2:32Selling continuous delivery externally
- 4:59High-quality health checks at scale
- 7:05RCB metrics effort and scalability
- 9:31Time windows for health checks
- 10:56Unit testing DB schema changes
- 12:42Automating rollbacks for DB changes
- 14:07Automating rollout validation
- 15:09Canaries and end-user impact
- 16:33Comparison thresholds limitations
- 17:56Zero-downtime DB migrations
- 19:43Migrations versus version control
- 20:47RCV runtime window choice
- 21:51False positives in health checks
- 23:13Culture for health check reliability
- 25:00Closing
related talks
19mLive Q&A
42mArchitecting for Reliability
28mReliability Lessons from Running High Traffic Services for More Than 10 Years
22mReliability as Code (RaC), A Startup Case Study
23mLive Q&A
23mDiving through RCAs a quest for reliability
27mContinuous Reliability. How?
45mApplication reliability with Azure Load Testing and Chaos Studio Q&A