Argo Workflows

Argo Workflows est un moteur d'orchestration de workflows natif Kubernetes, CNCF Graduated, développé par argoproj. Chaque step d'un workflow s'exécute dans un pod Kubernetes dédié. Il supporte deux modèles d'orchestration : steps (séquentiel ou parallèle) et dag (graphe acyclique dirigé). Conçu pour les pipelines CI/CD, le batch processing, les data pipelines et les workloads ML/AI. S'intègre naturellement avec les autres outils Argo (Events, Rollouts, CD).

Informations essentielles

Origine : Applatix → Intuit → CNCF Graduated  ·  Licence : Apache 2.0  ·  Architectures : x86_64, ARM64

Liens : Site officiel  ·  Documentation  ·  GitHub  ·  Releases

Support : CNCF Graduated. Communauté active, utilisé par Intuit, Alibaba, BlackRock en production.

Stack par défaut

Prérequis

Installation

kubectl create namespace argo

kubectl apply -n argo -f \
  https://github.com/argoproj/argo-workflows/releases/latest/download/install.yaml

# Vérifier l'installation
kubectl get pods -n argo
# workflow-controller et argo-server Running

Accéder à l'UI

kubectl port-forward svc/argo-server -n argo 2746:2746
# Ouvrir https://localhost:2746 (certificat auto-signé, accepter l'exception)

# Token de connexion
kubectl -n argo create token argo-server

CLI argo

curl -sLO \
  https://github.com/argoproj/argo-workflows/releases/latest/download/argo-linux-amd64.gz
gunzip argo-linux-amd64.gz
chmod +x argo-linux-amd64
sudo mv argo-linux-amd64 /usr/local/bin/argo

argo version

Workflows Steps (séquentiel / parallèle)

apiVersion: argoproj.io/v1alpha1
kind: Workflow
metadata:
  generateName: ci-pipeline-
  namespace: argo
spec:
  entrypoint: pipeline
  templates:
    - name: pipeline
      steps:
        - - name: test           # step 1 - séquentiel
            template: run-tests
        - - name: build          # step 2 - après test
            template: build-image
          - name: lint           # step 2bis - en parallèle avec build
            template: lint-code
        - - name: deploy         # step 3 - après build ET lint
            template: deploy-app

    - name: run-tests
      container:
        image: golang:1.22
        command: [go, test, ./...]

    - name: build-image
      container:
        image: gcr.io/kaniko-project/executor:latest
        args: [--context=., --destination=ghcr.io/org/app:latest]

    - name: lint-code
      container:
        image: golangci/golangci-lint:latest
        command: [golangci-lint, run]

    - name: deploy-app
      container:
        image: bitnami/kubectl:latest
        command: [kubectl, apply, -f, manifests/]

Workflows DAG (graphe de dépendances)

apiVersion: argoproj.io/v1alpha1
kind: Workflow
metadata:
  generateName: dag-pipeline-
  namespace: argo
spec:
  entrypoint: dag-tasks
  templates:
    - name: dag-tasks
      dag:
        tasks:
          - name: A
            template: run-task
            arguments:
              parameters: [{name: message, value: "A"}]
          - name: B
            dependencies: [A]
            template: run-task
            arguments:
              parameters: [{name: message, value: "B"}]
          - name: C
            dependencies: [A]
            template: run-task
            arguments:
              parameters: [{name: message, value: "C"}]
          - name: D
            dependencies: [B, C]
            template: run-task
            arguments:
              parameters: [{name: message, value: "D"}]

    - name: run-task
      inputs:
        parameters:
          - name: message
      container:
        image: alpine
        command: [echo, "{{inputs.parameters.message}}"]

WorkflowTemplate - templates réutilisables

apiVersion: argoproj.io/v1alpha1
kind: WorkflowTemplate
metadata:
  name: build-template
  namespace: argo
spec:
  templates:
    - name: build
      inputs:
        parameters:
          - name: image
      container:
        image: gcr.io/kaniko-project/executor:latest
        args: [--context=., --destination={{inputs.parameters.image}}]

# Soumettre un workflow depuis un template
argo submit --from workflowtemplate/build-template \
  -p image=ghcr.io/org/app:latest -n argo

CronWorkflow - planification

apiVersion: argoproj.io/v1alpha1
kind: CronWorkflow
metadata:
  name: nightly-backup
  namespace: argo
spec:
  schedule: "0 2 * * *"
  timezone: "Europe/Paris"
  workflowSpec:
    entrypoint: backup
    templates:
      - name: backup
        container:
          image: alpine
          command: [sh, -c, "echo 'backup started'"]

Mise à jour

kubectl apply -n argo -f \
  https://github.com/argoproj/argo-workflows/releases/latest/download/install.yaml

Troubleshooting

# État des pods
kubectl get pods -n argo

# Lister les workflows
argo list -n argo

# Logs d'un workflow
argo logs <workflow-name> -n argo --follow

# Workflow bloqué en Pending
argo get <workflow-name> -n argo
kubectl describe pod <pod-name> -n argo

# Logs du contrôleur
kubectl logs -n argo -l app=workflow-controller --tail=100

Commandes utiles

argo submit workflow.yaml -n argo         # Soumettre un workflow
argo list -n argo                         # Lister les workflows
argo get <nom> -n argo                    # Détails d'un workflow
argo logs <nom> -n argo -f                # Logs en temps réel
argo delete <nom> -n argo                 # Supprimer un workflow
argo retry <nom> -n argo                  # Rejouer un workflow échoué
argo stop <nom> -n argo                   # Arrêter un workflow en cours

Ressources

• Documentation : https://argo-workflows.readthedocs.io/
• GitHub : https://github.com/argoproj/argo-workflows
• Exemples : https://github.com/argoproj/argo-workflows/tree/master/examples
• Releases : https://github.com/argoproj/argo-workflows/releases