To Go or Not To Go: 4 Error Metrics for Confident Release Validation
summary
Spinnaker Kayenta integrates OverOps error fingerprints into canary configs to compare baseline and experiment across intervals using soak, traffic shadowing or 1:1 traffic, and it demonstrates a working webhook loop to rollback if production monitoring finds problems. Four actionable metrics drive pass/fail decisions: new errors, critical errors, resurfaced, and unique error total, measured by bytecode dynamic analysis (not just logs or APM). The demo shows real‑time canary reports plus deep root cause (variable state and commit links) to speed correct rollbacks and patches.
AI-written summary · about
chapters (5)
related talks
42mMetrics Driven Blue-green Deployments using Spinnaker’s Cloud...
40mCanarying Well: Lessons Learned from Canarying Large Populations
13mBeyond observability: Using reliability scores to drive results
22mFear of the computer
34mIdentifying Root Cause via Playback Metrics
11mKeynote: Smarter Progressive Delivery at Intuit
38mDreaming of high performance- a story about DORA metrics and Go
43mThe Invisible Door: Reliability Gaps in the Front End