Container or workflow orchestration?
| Dimension | Container orchestration | Workflow orchestration |
|---|---|---|
| What it orchestrates | Where long-running containers run | The order of steps in a process |
| Unit of work | A pod or task that should stay up | A step that runs and finishes |
| When something fails | Restart or reschedule the container | Retry the step, branch, or compensate |
| Lifetime | Always on | Seconds to years per run |
| Examples | Kubernetes, ECS, Nomad | Step Functions, Temporal, Airflow |
Rule of thumb
Both get called “orchestration”, which causes endless confusion. One keeps servers alive; the other keeps a business process moving.
Sources
- Kubernetes docs: Deployments
- Kubernetes docs: Pod lifecycle (container restarts)
- AWS docs: Amazon ECS services (replace failed tasks)
- Nomad docs: reschedule block
- AWS docs: Handling errors in Step Functions workflows (Retry and Catch)
- AWS docs: Step Functions quotas (1-year maximum execution time)
- Temporal docs: Workflow Execution (runs for seconds or years)
- Apache Airflow docs: Tasks (retries)