Release a cluster - E-MapReduce - Alibaba Cloud Documentation Center

Releasing a cluster is irreversible. All data on the cluster is permanently deleted and cannot be recovered.

Prerequisites

For a pay-as-you-go cluster, ensure that it is in the Creating, Running, or Idle state.
For a subscription cluster, ensure it is in the Running or Startup Failed state and contains no ECS instances. To release an ECS instance, see Release an instance.
Ensure that release protection is disabled for the cluster. For more information, see Enable and disable release protection.

Precautions

Before you can release a subscription EMR cluster, you must release all of its ECS instances, including those that are stopped but not yet released.
Note
After all ECS instances are released, the system automatically releases the EMR cluster within 15 minutes.
If you release an ECS instance while a task is running on it, the task fails.
If you release the ECS instances of a subscription EMR cluster before its expiration, the system refunds you for the unused portion of the subscription.
To view your bills, log on to the Expenses and Costs Console and choose Bills > Bill Details. For more information about refund rules and procedures, see Refund policy.
After you release an EMR cluster, log on to the ECS console and verify that the underlying ECS instances and disks are completely deleted. Do not rely only on the cluster status in the EMR console. Even if the cluster is shown as released, ECS instances or disks that remain can continue to incur charges. These charges are mainly for ECS instances and disks, plus a small amount of EMR pay-as-you-go resources. To verify when billing stops, check Bill Details in the Expenses and Costs console to confirm the final billing time and identify cloud resources that are still incurring charges.

Procedure

Log on to the E-MapReduce console.
- In the top navigation bar, select a region and resource group based on your requirements.
  - Region: The cluster is created in the selected region. You cannot change the region after the cluster is created.
  - Resource Group: By default, all resources in your account are displayed.
On the EMR on ECS page, find the cluster that you want to release and choose > Release in the Actions column.
Alternatively, go to the Basic Information page of the cluster. In the upper-right corner, choose All Operations > Cluster Operation > Release.
In the Release Cluster dialog box, click OK.

FAQ

How can I save a cluster's configuration before releasing it so that I can create a new cluster with the same configuration later?

Releasing a cluster permanently deletes its data and configuration, which cannot be recovered. To reuse the configuration, create a cluster template before releasing. For more information, see Create a cluster template.

Do EMR on ECS and StarRocks clusters support stopping the cluster to implement a cool-down period before deletion?

Currently, EMR on ECS and StarRocks clusters do not support stopping operations. As a result, a cool-down period before deletion cannot be implemented. Before you release a cluster, we recommend that you confirm that your business has stopped and your data has been backed up. Otherwise, releasing the cluster directly may interrupt running tasks or cause data loss.

Is the cluster release API call synchronous or asynchronous?

At the API level, the release cluster API call typically returns a synchronous response, such as HTTP 200. However, the actual release process on the backend is a long-running asynchronous task. Do not rely only on the API response to determine whether the release is complete. Instead, poll the cluster status by using the console or the DescribeCluster API operation to confirm that the release is complete.

What do I do if a cluster release API call times out or the request does not take effect?

The release request may not take effect due to network issues or delays in backend processing. We recommend that you:

Record the request ID of the release API call for troubleshooting. If you cannot obtain the request ID, you can manually release the cluster in the console or retry by using another API operation.
If you use a workflow scheduler such as Airflow, we recommend that you use CreateClusterFromTemplateOperator to dynamically create clusters and use TearDownClusterOperator to automatically release the clusters after tasks are complete. This ensures that resources are reclaimed in time and prevents resources from remaining after tasks are complete.

How can I reduce costs when pay-as-you-go instances in an EMR on ECS cluster are idle overnight?

EMR does not support stopping all components with one click. Before you stop the ECS instances, manually stop all EMR component services. Then, in the ECS console, stop the ECS instances one by one and select Economical Mode to reduce some costs. This helps you avoid continued charges for pay-as-you-go instances that are idle overnight.