Kubernetes Integration

Service accounts, RBAC, scheduling, storage and security — the Kubernetes foundations that make Argo Workflows reliable.

How Argo Workflows maps onto Kubernetes

Argo Workflows does not hide Kubernetes; it uses it directly. Workflows are custom resources, steps are pods, permissions are RBAC rules and configuration lives in ConfigMaps. Understanding which Kubernetes object is involved in a problem is usually enough to find the fix.

RBAC quick reference

WhoNeedsTypical grant
Workflow pods (executor)Report outputscreate, patch on workflowtaskresults
Resource templatesCreate objects such as JobsVerbs on that resource, on a separate service account
DevelopersSubmit and inspect workflowsCRUD on workflows, read on templates, pods/log
Platform adminsManage shared templatesCRUD on clusterworkflowtemplates

Multi-tenant layout

Run the controller and server in an argo namespace owned by the platform team, and give every team its own namespace with a dedicated workflow service account, a ResourceQuota, a LimitRange and, if needed, its own artifact repository. Pair this with SSO in the Argo Server so each user only sees the namespaces they belong to.

Production checklist

  • Requests and memory limits on every container — including executor containers via the controller ConfigMap.
  • Workload identity (IRSA, Workload Identity, Entra Workload ID) instead of static keys.
  • Restricted Pod Security Admission on workflow namespaces.
  • Network policies that still allow egress to the API server and artifact store.
  • Cost labels on every pod through podMetadata.

Read the complete Kubernetes integration guide, generate RBAC and quotas with the Kubernetes YAML generator, and score your manifests with the configuration checker.

Guides

Recommended tools