- Release Notes and Announcements
- Product Introduction
- Purchase Guide
- Getting Started
- EMR on CVM Operation Guide
- Planning Cluster
- Configuring Cluster
- Managing Cluster
- Instance Information
- Node Specification Management
- Checking and Updating Public IP
- Logging in to Clusters
- Cluster Scale-Out
- Cluster Scale-in
- Releasing Now
- Auto Scaling
- Graceful Scale-In
- Changing Configurations
- Automatic Replacement
- Exporting Software Configuration
- Cluster Scripts
- Cluster Termination
- Operation Logs
- Task Center
- Disk Management
- Managing Service
- Monitoring and Alarms
- Cluster Overview
- Node Status
- Service Status
- Cluster Event
- Log
- Application Analysis
- Cluster Inspection
- Java Analysis
- Monitoring Metrics
- Node Monitoring Metrics
- HDFS Monitoring Metrics
- YARN Monitoring Metrics
- ZooKeeper Monitoring Metrics
- HBase Monitoring Metrics
- Hive Monitoring Metrics
- Spark Monitoring Metrics
- Presto Monitoring Metrics
- Trino Monitoring Metrics
- ClickHouse Monitoring Metrics
- Druid Monitoring Metrics
- Kudu Monitoring Metrics
- Alluxio Monitoring Metrics
- PrestoSQL Monitoring Metrics
- Impala Monitoring Metrics
- Ranger Monitoring Metrics
- COSRanger Monitoring Metrics
- Doris Monitoring Metrics
- Kylin Monitoring Metrics
- Zeppelin Monitoring Metrics
- Oozie Monitoring Metrics
- Storm Monitoring Metrics
- Livy Monitoring Metrics
- Kyuubi Monitoring Metrics
- StarRocks Monitoring Metrics
- Kafka Monitoring Metrics
- Alarm Configurations
- Alarm Records
- EMR on TKE Operation Guide
- EMR Serverless HBase Operation Guide
- EMR Development Guide
- Hadoop Development Guide
- HDFS Common Operations
- HDFS Federation Management Development Guide
- HDFS Federation Management
- Submitting MapReduce Tasks
- Automatically Adding Task Nodes Without Assigning ApplicationMasters
- YARN Task Queue Management
- Practices on YARN Label Scheduling
- Hadoop Practical Tutorial
- Using API to Analyze Data in HDFS and COS
- Dumping YARN Job Logs to COS
- Spark Development Guide
- Hbase Development Guide
- Phoenix on Hbase Development Guide
- Hive Development Guide
- Presto Development Guide
- Sqoop Development Guide
- Hue Development Guide
- Oozie Development Guide
- Flume Development Guide
- Kerberos Development Guide
- Knox Development Guide
- Alluxio Development Guide
- Kylin Development Guide
- Livy Development Guide
- Kyuubi Development Guide
- Zeppelin Development Guide
- Hudi Development Guide
- Superset Development Guide
- Impala Development Guide
- Druid Development Guide
- TensorFlow Development Guide
- Kudu Development Guide
- Ranger Development Guide
- Kafka Development Guide
- Iceberg Development Guide
- StarRocks Development Guide
- Flink Development Guide
- Hadoop Development Guide
- Practical Tutorial
- API Documentation
- FAQs
- Service Level Agreement
- Contact Us
- Release Notes and Announcements
- Product Introduction
- Purchase Guide
- Getting Started
- EMR on CVM Operation Guide
- Planning Cluster
- Configuring Cluster
- Managing Cluster
- Instance Information
- Node Specification Management
- Checking and Updating Public IP
- Logging in to Clusters
- Cluster Scale-Out
- Cluster Scale-in
- Releasing Now
- Auto Scaling
- Graceful Scale-In
- Changing Configurations
- Automatic Replacement
- Exporting Software Configuration
- Cluster Scripts
- Cluster Termination
- Operation Logs
- Task Center
- Disk Management
- Managing Service
- Monitoring and Alarms
- Cluster Overview
- Node Status
- Service Status
- Cluster Event
- Log
- Application Analysis
- Cluster Inspection
- Java Analysis
- Monitoring Metrics
- Node Monitoring Metrics
- HDFS Monitoring Metrics
- YARN Monitoring Metrics
- ZooKeeper Monitoring Metrics
- HBase Monitoring Metrics
- Hive Monitoring Metrics
- Spark Monitoring Metrics
- Presto Monitoring Metrics
- Trino Monitoring Metrics
- ClickHouse Monitoring Metrics
- Druid Monitoring Metrics
- Kudu Monitoring Metrics
- Alluxio Monitoring Metrics
- PrestoSQL Monitoring Metrics
- Impala Monitoring Metrics
- Ranger Monitoring Metrics
- COSRanger Monitoring Metrics
- Doris Monitoring Metrics
- Kylin Monitoring Metrics
- Zeppelin Monitoring Metrics
- Oozie Monitoring Metrics
- Storm Monitoring Metrics
- Livy Monitoring Metrics
- Kyuubi Monitoring Metrics
- StarRocks Monitoring Metrics
- Kafka Monitoring Metrics
- Alarm Configurations
- Alarm Records
- EMR on TKE Operation Guide
- EMR Serverless HBase Operation Guide
- EMR Development Guide
- Hadoop Development Guide
- HDFS Common Operations
- HDFS Federation Management Development Guide
- HDFS Federation Management
- Submitting MapReduce Tasks
- Automatically Adding Task Nodes Without Assigning ApplicationMasters
- YARN Task Queue Management
- Practices on YARN Label Scheduling
- Hadoop Practical Tutorial
- Using API to Analyze Data in HDFS and COS
- Dumping YARN Job Logs to COS
- Spark Development Guide
- Hbase Development Guide
- Phoenix on Hbase Development Guide
- Hive Development Guide
- Presto Development Guide
- Sqoop Development Guide
- Hue Development Guide
- Oozie Development Guide
- Flume Development Guide
- Kerberos Development Guide
- Knox Development Guide
- Alluxio Development Guide
- Kylin Development Guide
- Livy Development Guide
- Kyuubi Development Guide
- Zeppelin Development Guide
- Hudi Development Guide
- Superset Development Guide
- Impala Development Guide
- Druid Development Guide
- TensorFlow Development Guide
- Kudu Development Guide
- Ranger Development Guide
- Kafka Development Guide
- Iceberg Development Guide
- StarRocks Development Guide
- Flink Development Guide
- Hadoop Development Guide
- Practical Tutorial
- API Documentation
- FAQs
- Service Level Agreement
- Contact Us
Overview
After graceful scale-in is enabled, if the scale-in action is triggered when a node is executing tasks, the node will not be released immediately. Instead, it will be released after completing the tasks. Graceful scale-in is available for both automatic scaling and manual scale-in.
Note
YARN, HBase, and Presto (renamed Trino in EMR v2.7.0 and v3.40 or later) support graceful scale-in. However, Presto cannot be scaled in gracefully in Ranger, Kerberos, and OpenLDAP integration scenarios.
Directions
Auto-scaling
Auto-scaling allows you to enable or disable graceful scale-in for all scale-in rules. This feature is called global graceful scale-in and is disabled by default. When you add or edit a single scale-in rule, graceful scale-in is enabled by default, and the default duration is 60 seconds (valid range: 60–1800 seconds).
Note
When global graceful scale-in and a single rule are enabled, graceful scale-in takes effect for the rule.
1. Log in to the EMR console, click the ID/name of the target cluster to enter the cluster details page, and click Auto-scaling.
2. In the Scaling rule management section on the Auto-scaling page, click Add rule and add a scale-in rule.
Manual scale-in
When you try to manually remove a node, graceful scale-in is disabled by default. When you enable this feature, the default duration is 60 seconds (valid range: 60–1800 seconds).
1. Click the ID/name of the target cluster to enter the cluster details page. Then, select Cluster resources > Resources.
2. Select the node to remove and click Scale in. In this case, graceful scale-in is disabled by default. You can enable it and set a duration.
3. After finishing your settings, click Next, confirm the node information, and click Start termination.
Yes
No
Was this page helpful?