Container Argo workflow clusters are serverless Argo Workflows environments built on Alibaba Cloud Container Compute Service (ACS). Instead of provisioning and managing a Kubernetes cluster, you get a fully managed workflow engine that scales automatically, optimizes costs with Best Effort instances, and stays fully compatible with the open-source Argo Workflows project.
Use cases
Argo Workflows is a cloud-native workflow engine and a graduated project of the Cloud Native Computing Foundation (CNCF). This status indicates that the project meets the highest standards for user adoption, security, and widespread use. Each workflow step runs in a Kubernetes pod, making it lightweight, container-native, and capable of running thousands of jobs in parallel without virtual machine (VM) overhead.
Container Argo workflow clusters are a good fit when:
You run large-scale batch or ML workloads. The cluster schedules thousands of parallel pods across multiple zones and releases resources immediately when jobs finish — no idle capacity to pay for.
You need compute-intensive pipelines with minimal ops overhead. Submit a workflow and the cluster handles scheduling, scaling, and zone failover. You don't manage Kubernetes node groups or version upgrades.
You orchestrate heterogeneous jobs. Argo Workflows natively supports standard jobs alongside Spark, Ray, and Tensor jobs in a single directed acyclic graph (DAG).
You're running CI/CD or infrastructure automation. Each pipeline step is an isolated container, giving you reproducible, auditable runs.
You already use Argo Workflows on ACK or another Kubernetes cluster. Workflow clusters are fully compatible with open-source Argo Workflows — migrate existing workflows without modification.
It is widely used in industries such as autonomous driving, scientific computing, financial quantization, and digital media.
How it works
Container Argo workflow clusters are serverless workflow engines hosted on Kubernetes clusters. The key components are:
ACS pods — the compute units that run your workflow steps. Each pod runs as a container inside ACS, consuming one IP address from a vSwitch. Pods are created on demand and released when the step finishes.
Enhanced control plane — an optimized layer that improves scheduling throughput, stability, and observability beyond the open-source defaults.
Key features
| Feature | What it gives you |
|---|---|
| Open-source compatible | Fully compatible with open-source Argo Workflows. Migrate existing workflows from ACK clusters or any Kubernetes cluster without modifying a single YAML file. |
| No O&M overhead | No Kubernetes cluster to manage. Version upgrades, control plane patches, and infrastructure maintenance are handled for you. |
| Automatic scaling | Resources scale out when workflows start and are released the moment steps finish. Best Effort instances keep costs low. |
| Multi-zone reliability | Multi-zone load balancing for highly reliable scheduling. The workflow engine instance automatically selects a zone with sufficient inventory to create ACS pods. If a zone has insufficient inventory, the workflow cannot retrieve ACS pod resources and fails. |
| Enhanced control plane | Improved scheduling performance, higher throughput, better stability, and built-in observability — all tuned for large-scale workflow workloads. |
| Enhanced OSS storage | Supports large file uploads, streaming, and artifact garbage collection (GC) for workflows that produce or consume significant data volumes. |
| Community expert support | Access to Argo Workflows community experts who can help optimize your workflow design for performance and cost. |
Available regions
Container Argo workflow clusters are available in the following regions:
| Region name | Region ID |
|---|---|
| China (Beijing) | cn-beijing |
| China (Zhangjiakou) | cn-zhangjiakou |
| China (Hangzhou) | cn-hangzhou |
| China (Shanghai) | cn-shanghai |
| China (Shenzhen) | cn-shenzhen |
| China (Heyuan) | cn-heyuan |
| China (Guangzhou) | cn-guangzhou |
| China (Hong Kong) | cn-hongkong |
| Singapore | ap-southeast-1 |
| Malaysia (Kuala Lumpur) | ap-southeast-3 |
| Indonesia (Jakarta) | ap-southeast-5 |
| Thailand (Bangkok) | ap-southeast-7 |
| Japan (Tokyo) | ap-northeast-1 |
| Germany (Frankfurt) | eu-central-1 |
| UK (London) | eu-west-1 |