Help Center/ MapReduce Service/ User Guide/ Managing Clusters/ Managing MRS Cluster Nodes/ Starting or Stopping All Roles on an MRS Cluster Node
Updated on 2026-10-09 GMT+08:00
View PDF
Share

Starting or Stopping All Roles on an MRS Cluster Node

Starting or stopping roles on an MRS cluster node helps cluster administrators and O&M engineers maintain and troubleshoot nodes.

During routine O&M, an MRS node may need to be taken offline for maintenance due to hardware aging, disk faults, or OS patch upgrades. Before performing maintenance, stop all role instances on the node to ensure service processes exit safely. After maintenance is complete, start all role instances to restore services on the node.

  • Node fault maintenance: If a host encounters a hardware exception, disk fault, or system issue, stop all roles on the node before performing maintenance. After rectifying the fault, start the roles to restore services.
  • Rolling upgrades and maintenance: When cluster configurations change, stop roles node by node to perform maintenance without interrupting entire cluster services.
  • Temporary resource release: During off-peak hours, temporarily stop roles on some nodes to save resources. Start the roles before peak hours.

    Stopping all roles on a node interrupts all service processes running on the node and may cause active tasks to fail. Before proceeding, evaluate the impact on services and perform this operation during off-peak hours. Before stopping any roles, check whether critical tasks are running on the node and migrate them if necessary.

Prerequisites

  • The IAM users have been synchronized in advance. You can do this by clicking Synchronize next to IAM User Sync on the Dashboard page of the cluster details.
  • You have logged in to MRS Manager. For how to log in, see Accessing MRS Manager.
  • Before stopping a role, check the status of role instances and running tasks on MRS Manager and evaluate how the operation will affect services.
  • For roles deployed in active/standby mode (such as NameNode or ResourceManager), perform an active/standby switchover before stopping an active role instance to prevent service interruption.

Starting or Stopping Roles

You can start or stop roles using either the management console or MRS Manager.

  1. Log in to the MRS console.
  2. On the Active Clusters page, select a running target cluster and click its name to go to the cluster details page.
  3. On the cluster details page, click Nodes.
  4. Unfold the node group information and select the check box of the target node.
  5. Choose Node Operation > Start All Roles or Stop All Roles to perform the required operation.
  1. Log in to FusionInsight Manager of the MRS cluster.

    For details about how to log in to FusionInsight Manager, see Accessing MRS Manager.

  2. Click Hosts to go to the host list page.

    For clusters of MRS 2.x or earlier, click Hosts.

  3. Select the check box of the target hosts.
  4. Select Start All Instances or Stop All Instances from the More drop-down list to start or stop all role instances.

    In the confirmation dialog box that appears, enter the current user password and click OK.

    After the start or stop operation completes, verify the result as follows:

    1. On MRS Manager, choose Hosts and click the target host name to open the host details page.
    2. Click the Process tab and check the running status of each role instance.

FAQs

Q: What happens to running tasks when a role on a node is stopped?

A: Stopping a role terminates all running containers and processes on the node. YARN tasks (such as MapReduce and Spark jobs) are automatically rescheduled to other healthy nodes if the tasks support retries and have not exceeded the maximum retry limit. Regions on an HBase RegionServer automatically migrate to other nodes, which may cause brief read/write delays. Before stopping a role, check running tasks on the node in YARN ResourceManager and stop them if necessary.

Q: What should I do if some roles fail to start after a batch start operation?

A: View the logs for the failed role instance in the instance list on the host details page of MRS Manager. Common causes include insufficient disk space, port conflicts, configuration errors, or unready dependent roles. After resolving the issue, click the start button next to the failed instance to start it individually.

Q: Can I safely stop or restart a host after stopping all its roles?

A: Yes. After stopping all roles and confirming that every instance has stopped completely, you can safely stop or restart the host. Do not stop a host while roles are still running. Doing so may cause data corruption or service exceptions.

Q: What is the difference between stopping a role and isolating a node?

A: Stopping a role only terminates specific service processes on a node. The node remains under cluster management and resumes normal operations once started. Isolating a node stops all its roles and removes the node from cluster scheduling so no new tasks are assigned to it. For long-term host maintenance, stop all roles on the node before performing node isolation.

Helpful Links

Parent Topic: Managing MRS Cluster Nodes