1,000,000 members · Thursday 14:30
14:30:00
without argo rollouts deploying
error rate %
0.2%
— slo threshold 1%
⚠ 80,000 members affected · 18 min degraded service
with argo rollouts canary 0%
→setWeight 10%
⏸pause 2 minutes
⚡analysis — kube_pod_status_ready
→setWeight 30%
analysisrun
querying prometheus…
✓ 0 members affected · rollback fired automatically
THE SAME DEPLOY. TWO OUTCOMES.
Argo Rollouts intercepted the bad release at 10% canary. The AnalysisRun detected unready pods and fired the rollback before the pause timer even expired.
Without
80,000
members hit the broken version
18 min
degraded service
With
0
members hit the broken version
< 3 min
canary caught and reverted
✓10% canary weight — 90% of members always on stable
✓AnalysisRun queries Prometheus every 30s — no human required
✓Rollback fires in seconds, not 18 minutes
env: production
phase: 5b — progressive delivery