tencent cloud

Elastic MapReduce

Release Notes and Announcements
Release Notes
Announcements
Security Announcements
Product Introduction
Overview
Strengths
Architecture
Features
Use Cases
Constraints and Limits
Technical Support Scope
Product release
Purchase Guide
EMR on CVM Billing Instructions
EMR on TKE Billing Instructions
EMR Serverless HBase Billing Instructions
Getting Started
EMR on CVM Quick Start
EMR on TKE Quick Start
EMR on CVM Operation Guide
Planning Cluster
Administrative rights
Configuring Cluster
Managing Cluster
Managing Service
Monitoring and Alarms
TCInsight
EMR on TKE Operation Guide
Introduction to EMR on TKE
Configuring Cluster
Cluster Management
Service Management
Monitoring and Ops
Application Analysis
EMR Serverless HBase Operation Guide
EMR Serverless HBase Product Introduction
Quotas and Limits
Planning an Instance
Managing an Instance
Monitoring and Alarms
Development Guide
EMR Development Guide
Hadoop Development Guide
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
JupyterLab Development Guide
MLflow Development Guide
Practical Tutorial
Practice of EMR on CVM Ops
Data Migration
Practical Tutorial on Custom Scaling
API Documentation
History
Introduction
API Category
Cluster Resource Management APIs
Cluster Services APIs
User Management APIs
Data Inquiry APIs
Scaling APIs
Configuration APIs
Other APIs
Serverless HBase APIs
YARN Resource Scheduling APIs
Making API Requests
Data Types
Error Codes
FAQs
EMR on CVM
Service Level Agreement
Contact Us
ドキュメントElastic MapReduce EMR on CVM Operation GuideTCInsightConfiguration Center - Identification and Diagnostics

Configuration Center - Identification and Diagnostics

PDF
フォーカスモード
フォントサイズ
最終更新日: 2026-01-13 18:04:48

Feature Introduction

The Configuration Center is a centralized management feature designed to help enterprises efficiently manage configurations and policies in big data cluster environments. Through the Configuration Center, users can flexibly adjust and optimize policies for core modules, including basic diagnosis, computing insights, storage insights, and resource insights, thereby achieving comprehensive regulatory analysis of big data clusters.
The Configuration Center supports the following key features:
Basic Diagnosis Policies: Provides multiple AI feature recognition models (including full load, data spike, and mean offset) and predictive analysis capabilities.
Computing Insight Policies: Supports full lifecycle computing insight values configuration for engines in the Hadoop ecosystem, including Spark, Hive, YARN, and Trino.
Storage Insight Policies: Analyzes file and Hive data tables for HDFS and COS storage, enabling identification of large and small files, as well as the configuration of cold and hot data classifications.
Resource Insight Policies: Analyzes the usage of physical and virtual resources by computing cluster, engine queue, and component dimension.

Operation Steps

1. Log in to the Tencent Cloud EMR Console > TCInsight > Configuration Center, select TCInsight > Configuration Center in the left sidebar on the console, and click to go to the Configuration Center page.
2. On the Configuration Center page, select the region availability zone and cluster you want to configure.
3. You can adjust related dimensional policy parameters and status as needed.

Configuration Center Policy Details

Basic Diagnosis Policies are as follows:
Dimension
Feature
Metric
Trigger Policy
Default Status
Severity Level
Whether Disablement Is Supported
Basic Diagnosis
Data spike
HBase RS request processing latency
Feature analysis
Enabled
Low
Yes
HBase read-write total request volume
Enabled
Yes
HBase RS slow operation count_slowAppendCount
Enabled
Yes
HBase RS slow operation count_slowDeleteCount
Enabled
Yes
HBase RS slow operation count_slowPutCount
Enabled
Yes
Percentage of used node memory
Enabled
Yes
TCP LISTEN exception_ListenDrops
Enabled
Yes
TCP retransmission rate_InErrRate
Enabled
Yes
SR EDITLOG write latency
Enabled
Yes
SR FE query latency
Enabled
Yes
Full load
HDFS storage space utilization
threshold=90
Enabled
Yes
Percentage of used HiveServer2 heap memory
threshold=90
Enabled
Yes
Node storage space utilization
threshold=90
Enabled
Yes
TCP socket memory
threshold=3221225472
Enabled
Yes
UDP socket memory
threshold=3221225472
Enabled
Yes
TCP4 connection status_CloseWait
threshold=50000
Enabled
Yes
TCP4 connection status_ESTABLISHED
threshold=50000
Enabled
Yes
TCP4 connection status_TimeWait
threshold=50000
Enabled
Yes
YARN Node Manager heap memory utilization
threshold=90
Enabled
Yes
YARN Resource Manager heap memory utilization
threshold=90
Enabled
Yes
Mean offset
TCP4 connection count_CLOSE-WAIT
Feature analysis
Enabled
Yes
TCP4 connection count_ESTABLISHED
Enabled
Yes
TCP4 connection count_TIME-WAIT
Enabled
Yes
Number of system processes
Enabled
Yes
Computing Insight Policies are as follows:
Dimension
Engine
Insight Item
Severity Level
Computing Insights
Hive
JOIN data inflation
Low
Empty input
Low
Full table scanning of partitioned tables
Low
Small input files
Medium
Excessive data scanning
High
MapJoin optimization
Low
Excessive metadata scanning
High
Large table scanning
High
Prolonged compilation duration
High
Improper parameter
Medium
Spark
BroadcastJoin optimization
Low
Query condition not pushed down
Low
CPU resource wastage
Low
JOIN data inflation
Low
Data skew
High
Empty Task input
High
ExecutorGC
Medium
Full table scanning of partitioned tables
Low
Global sorting
High
Too many small input files
High
Insufficient resources
High
Excessive data scanning
High
Peak memory exceeds limit
Low
Memory resource wastage
Low
Too many small output files
High
Task reading lag
High
Abnormal resource overhead
High
Scheduling delay
Low
ScheduleOverhead
High
Scheduling skew
High
ShuffleFailure
Medium
Slow Tasks
Medium
Too small Task input data
High
Abnormal Stage duration
Medium
StageScheduleDelay
Medium
Multiple Spark apps concurrently insert into the same table
Low
ShuffleServer application writes TopN
Low
Trino
Full table scanning of partitioned tables
Low
Excessive data scanning
High
StarRocks
JOIN data inflation
Low
Data Skew
High
Full table scanning of partitioned tables
High
Excessive data scanning
High
Storage Insight Policies are as follows:
Dimension
Type
Monitoring Data
Trigger Policy
Default Status
Whether Configuration Is Supported
HDFS Storage Insights
Large and small files
Large file storage greater than 3072 MB
Percentage of storage capacity
Greater than 30%
Yes
Small file storage greater than 0 MB but less than 2 MB
Percentage of file count
Greater than 30%
Yes
Empty file storage equal to 0 MB
Percentage of file count
Greater than 15%
Yes
Junk directory last modified 7 days ago
The junk directory matches the regular expression
.*/warehouse/.*/_temporary/.*/task_.*|.*/.hive-staging.*/
No
Cold and hot data
Last access time of hot files
Date
Less than 1 month
Yes
Last access time of warm files
Date
Greater than or equal to 1 month, less than or equal to 1 year
Yes
Last access time of cold files
Storage capacity
An alert is triggered when the percentage of storage capacity exceeds 50%.
Yes
StarRocks Insight
Data Tables
Bucketing skew
Bucketing stored amount
Bucketing stored amount skew deviation exceeds 1%
Yes
Index Optimization
Non-index queries/Number of queries
Queries with non-index ratio of 1%
Query count/day threshold: 2
Yes
Resource Insight Policies are as follows:
Dimension
Type
Insight Item
Severity Level
Insight Default Rule (Configurable)
Default Status
Whether Disablement Is Supported
Resource Insights
Cluster
Cluster resource CPU sustained idle
Low
Maximum idle utilization 10%
Duration: 30 minutes
Disabled
Yes
Cluster resource CPU sustained full load
High
Minimum full load utilization 90%
Duration: 30 minutes
Enabled
Yes
Cluster resource memory sustained idle
Low
Maximum idle utilization 10%
Duration: 30 minutes
Disabled
Yes
Cluster resource memory sustained full load
High
Minimum full load utilization 90%
Duration: 30 minutes
Enabled
Yes
Yarn queue
Yarn queue physical CPU overused
High
CPU over-allocation ratio 10%
Cluster overall CPU utilization greater than 90%
Enabled
Yes
Yarn queue virtual CPU wasted
Low
CPU wastage ratio 10%
Cluster overall CPU utilization greater than 90%
Disabled
Yes
Yarn queue physical memory overused
High
Memory over-allocation ratio 10%
Cluster overall memory utilization greater than 90%
Enabled
Yes
Yarn queue virtual memory wasted
Low
Memory wastage ratio 10%
Cluster overall memory utilization greater than 90%
Disabled
Yes
Yarn queue virtual CPU sustained idle
Low
Maximum idle utilization 10%
Duration: 30 minutes
Disabled
Yes
Yarn queue virtual CPU sustained full load
High
Minimum full load utilization 90%
Duration: 30 minutes
Enabled
Yes
Yarn queue virtual memory sustained idle
Low
Maximum idle utilization 10%
Duration: 30 minutes
Disabled
Yes
Yarn queue virtual memory sustained full load
High
Minimum full load utilization 90%
Duration: 30 minutes
Enabled
Yes
StarRocks
StarRocksBe CPU sustained idle
Low
Maximum idle utilization 10%
Duration: 30 minutes
Disabled
Yes
StarRocksBe CPU sustained full load
High
Minimum full load utilization 90%
Duration: 30 minutes
Enabled
Yes
StarRocksBe memory sustained idle
Low
Maximum idle utilization 10%
Duration: 30 minutes
Disabled
Yes
StarRocksBe memory sustained full load
High
Minimum full load utilization 90%
Duration: 30 minutes
Enabled
Yes
StarRocksFe CPU sustained idle
Low
Maximum idle utilization 10%
Duration: 30 minutes
Disabled
Yes
StarRocksFe CPU sustained full load
High
Minimum full load utilization 90%
Duration: 30 minutes
Enabled
Yes
StarRocksFe memory sustained idle
Low
Maximum idle utilization 10%
Duration: 30 minutes
Disabled
Yes
StarRocksFe memory sustained full load
High
Minimum full load utilization 90%
Duration: 30 minutes
Enabled
Yes

ヘルプとサポート

この記事はお役に立ちましたか?

フィードバック