Argo Workflows rewards a little discipline. These are the practices that keep it maintainable as a team and a codebase grow.
The practices
- Set limits and activeDeadlineSeconds on templates
- Use workflow GC (TTL) to clean up finished runs
- Store artifacts in object storage, not the cluster
What good looks like
A DAG workflow with dependencies.
apiVersion: argoproj.io/v1alpha1
kind: Workflow
metadata: { generateName: build- }
spec:
entrypoint: main
templates:
- name: main
dag:
tasks:
- { name: build, template: run, arguments: { parameters: [{ name: cmd, value: make build }] } }
- { name: test, template: run, dependencies: [build], arguments: { parameters: [{ name: cmd, value: make test }] } }
- name: run
inputs: { parameters: [{ name: cmd }] }
container: { image: alpine, command: [sh, -c], args: ['{{inputs.parameters.cmd}}'] }Anti-patterns to retire
- No resource limits on step containers
- Unbounded retries hammering the cluster
- Keeping completed workflows forever, bloating etcd
Want senior help putting Argo Workflows to work - or a team bootcamp on it? CodeWizard runs Israel's largest DevOps community and embeds named senior engineers with a transparent hours log. Book a free consult or explore the Academy.