Argo Workflows
Workflow orchestration component for Foundation automations and pipelines.
Component Category
Workflows and automation / workflow orchestration
Component Description
Argo Workflows is a Kubernetes-native workflow engine for running multi-step jobs, pipelines, and batch processes.
Why It Is Used
In BullSequana AI Foundation, Argo Workflows provides the execution layer for structured automation flows that need ordered steps, retries, artifacts, and Kubernetes-native scaling. It is a key building block for document pipelines, event-driven jobs, and operational automations.
Learn More
Deployment notes
Argo Workflows deploys into the argo-workflows namespace at sync wave 1 in the proai tier. Workflow history is persisted in the CNPG PostgreSQL cluster with archiving enabled. Completed workflow records are deleted after a 300-second TTL, and workflow pods are garbage-collected on completion. SSO is configured through Keycloak OIDC with RBAC enabled.
Workflow pods run as the argo-workflows-workflow service account. It can read (get and list) Secrets in the argo-workflows namespace and cannot write them. The chart also creates the Role that artifact garbage collection needs.
UI example image
The example templates in the Argo Workflows UI reference the public argoproj/argosay:v2 image. BullSequana AI ships a rebuild of that image in the platform registry, argo-workflows/argosay-bsq, which runs as a non-root user. A Kyverno policy, mutate-argo-example-image, rewrites exactly those upstream references on new workflow pods in the argo-workflows namespace to the platform registry image. The workflow controller knows the image's entrypoint without a registry lookup. The UI examples therefore run on clusters without access to Docker Hub.
The [example_image] table of the argo-workflows component sets the rewritten references and the replacement image's repository and tag. The policy changes only the pod: the Workflow object keeps the upstream reference. Workflows that reference other public images are not rewritten; point them at the platform registry. While Kyverno is unavailable, pods are admitted unchanged and pull the upstream image.
Interacts With
Argo Events, which can trigger workflows from external or internal events.Rook Ceph, which provides S3-compatible storage for workflow artifacts, inputs, and outputs through RGW.Temporal, which covers a different orchestration pattern for durable application workflows.Prometheus,Grafana, andGrafana Loki, through workflow metrics and logs used for operations and troubleshooting.Kyverno, which rewrites the UI example image on workflow pods to the platform registry.