Failed to Reduce Task Nodes

Issue

A user fails to scale in an MRS 2.x cluster by reducing the number of task nodes to 0 on the MRS console.

Symptom

When the number of task nodes in an MRS cluster is reduced on the MRS console, the following information is displayed:

This operation is not allowed because the number of instances of NodeManager will be less than the minimum configuration after scale-in, which may cause data loss.

Cause Analysis

The NodeManager service of the core node is stopped. If the number of task nodes is changed to 0, there will be no NodeManager in the cluster and the Yarn service will be unavailable. Therefore, MRS allows the reduction of task nodes only when the number of NodeManagers is greater than or equal to 1.

Procedure

1. Choose Services > Yarn > Instances.
   • For versions earlier than MRS 2.0.1: Log in to MRS Manager, and choose Services > Yarn > Instance.
   • For MRS 2.0.1 or later: Click the cluster name on the MRS console and choose Components > Yarn > Instances.
   • For MRS 3.x or later: Log in to FusionInsight Manager, choose Cluster, click the name of the target cluster, and choose Services > Yarn > Instance.

2. Select the NodeManager instance of the core node, click More, and select Start Instance.

3. Reduce the number of task nodes on the cluster details page.
   1. Click the cluster name, and select the Nodes tab.
   2. Locate the row that contains the task node group and click Scale In in the Operation column.
   3. Click OK. In the displayed dialog box, click Yes.

4. After the scale-in is successful, stop NodeManager of the core node if you do not need it.

Summary and Suggestions

You are advised not to stop NodeManager of the core node.