Tuning Argo Rollouts for Thousands of Workloads
summary
Adobe runs Argo Rollouts at scale: ~6,000 Rollout objects across clusters managing over 10k customer services with ~10k reconciliations average and up to 5k per cluster, using Canary deployments with automatic rollback and combining six metrics (error ratio, absolute errors, request volume, dual-deployment checks). Migration required a PR to add a scaleDown attribute, fixed Helm/immutable-secret conflicts with resourcePolicy keep plus cleanup jobs, and traded increased resource usage and slower deploys for far fewer full-production outages.
AI-written summary · about
chapters (7)
- 0:00Argo Rollouts at Adobe
- 3:18Adopting Argo Rollouts at scale
- 8:43Rollbacks with deployments
- 13:18Stability benefits of Argo Rollouts
- 15:09Migration challenges
- 19:28False positives in rollouts
- 21:39Key takeaways
related talks
06m⚡ Lightning Talk: Improving Release Reliability with Argo Rollouts Across 10k Customer Services
18mTailored Deployment Strategies: Argo Rollouts for Diverse Applications
30mArgo Rollouts at Scale: How We Brought Automated Rollback to 2100+ Micro-serv...
24mArgo Rollouts from the Trenches – Next Level Progressive Delivery Best Practices
30mProgressive Delivery Made Easy With Argo Rollouts
25mSailing Across the Ocean of Argo Ecosystem Like a Pro
22mReal Time Argo Rollouts Analytics Powered by Notification Engine
35mProgressive Delivery with Argo Rollouts