Agent Skills: Orchestrating Deployment Pipelines

Orchestrating Deployment Pipelines

Overview

Orchestrate multi-stage deployment pipelines that coordinate builds, tests, approvals, and releases across environments (dev, staging, production). Implement deployment strategies including blue-green, canary, rolling updates, and feature flags using Kubernetes, cloud-native services, and CI/CD platforms.

Prerequisites

• CI/CD platform configured (GitHub Actions, GitLab CI, Jenkins, ArgoCD)
• Kubernetes cluster with kubectl access or cloud deployment target (ECS, Cloud Run, App Engine)
• Container registry with built and tagged images ready for deployment
• Environment-specific configuration (secrets, environment variables) stored securely
• Monitoring and alerting configured to detect deployment failures

Instructions

1. Define the deployment topology: target environments, promotion flow (dev -> staging -> production), and approval gates
2. Select deployment strategy per environment: rolling update for staging, canary or blue-green for production
3. Generate deployment manifests (Kubernetes Deployments, Services, Ingress) or cloud service configurations
4. Implement pre-deployment checks: database migration status, dependency health, configuration validation
5. Configure canary analysis: route 5-10% of traffic to new version, monitor error rate and latency for 15 minutes before full rollout
6. Add post-deployment verification: smoke tests, health check endpoints, synthetic monitoring
7. Implement automated rollback triggers: revert if error rate exceeds 1% or P99 latency doubles during canary phase
8. Set up deployment notifications: Slack messages with deployment status, version, environment, and commit link
9. Document the deployment runbook with manual intervention procedures for edge cases

Output

• Deployment pipeline configurations (GitHub Actions workflows, ArgoCD Applications)
• Kubernetes manifests with deployment strategy annotations
• Canary analysis configuration (Flagger, Argo Rollouts)
• Pre/post-deployment hook scripts
• Deployment runbook with rollback procedures

Error Handling

| Error | Cause | Solution | |-------|-------|---------| | ImagePullBackOff | Image tag not found in registry or auth failure | Verify image exists with docker manifest inspect; check imagePullSecrets | | CrashLoopBackOff | Application failing to start in new version | Check pod logs with kubectl logs; verify environment variables and config maps | | Canary analysis failed | Error rate or latency exceeded threshold during canary | Automatic rollback triggered; investigate logs from canary pods before retrying | | Deployment stuck in Progressing | Insufficient resources or pod scheduling failure | Check kubectl describe deployment for events; verify resource requests and node capacity | | Database migration failed | Schema conflict or lock timeout | Run migrations independently before deployment; add retry logic and connection timeout |

Examples

• "Create a deployment pipeline that builds on PR merge, deploys to staging automatically, runs integration tests, then requires manual approval for production with canary rollout."
• "Set up Argo Rollouts for a Kubernetes deployment with 10% canary traffic, Prometheus-based analysis, and automatic rollback on error rate > 0.5%."
• "Generate a blue-green deployment for an ECS service with ALB target group switching and automatic rollback on health check failure."

Resources

• Kubernetes deployment strategies: https://kubernetes.io/docs/concepts/workloads/controllers/deployment/
• Argo Rollouts: https://argoproj.github.io/argo-rollouts/
• Flagger (progressive delivery): https://flagger.app/
• AWS ECS blue-green: https://docs.aws.amazon.com/AmazonECS/latest/developerguide/deployment-type-bluegreen.html