Don't Do What Donny Don't Does: The 10 Dos and 500 Don’ts of Workflows...
summary
Practical production rules for Argo Workflows: instrument the controller, API and pods with observability, set container requests/limits and tune requeue time, and use TTL/archiving instead of keeping CRs in-cluster. Avoid the built-in log archiver—use DataDog or Loki and add UI links. Use SSO and least-privilege service accounts; example uses Tailscale ingress for private Argo Server. At scale prefer one workflow controller replica, tune control plane or shard clusters, and expect gains like 6 hours to 30 minutes and 99.9% reliability and up to 2/3 cost reduction.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 1:24Argo Workflows overview
- 3:50Observability in Argo
- 7:00CPU and memory limits
- 10:10Workflow and cluster cleanup
- 11:36Security for Argo Workflows
- 14:47Kubernetes optimization
- 16:53Scaling Argo Workflows
- 21:24SDKs for workflows
- 22:06Real-world savings
related talks
20mDemystifying Argo Workflows: An Architectural Deep Dive
36mArgo Workflows: Intro, Update and Deep Dive
29mLarge Scale Batch Processing with Argo Work...
25mEvolving Argo Workflows for AI: New Features, Tuning, and Multi-Tenant Isolation
27mThe Life of an Argo Workflows Salesperson
05mArgo Workflows 4, What's New And What's Next
06mComparing Argo Workflows and Apache Airflow at Scale
27mBuilding a Data Science Platform with Argo Workflows