Yves-yuan / compass

Compass is a task diagnosis platform for bigdata

Compass

中文文档

Compass is a big data task diagnosis platform, which aims to improve the efficiency of user troubleshooting and reduce the cost of abnormal tasks for users.

The key features:

• Non-invasive, instant diagnosis, you can experience the diagnostic effect without modifying the existing scheduling platform.

• Supports multiple scheduling platforms(DolphinScheduler, Airflow, or self-developed etc.)

• Supports Spark 2.x or 3.x, Hadoop 2.x or 3.x troubleshooting.

• Supports workflow layer exception diagnosis, identifies various failures and baseline time-consuming abnormal problems.

• Supports Spark engine layer exception diagnosis, including 14 types of exceptions such as data skew, large table scanning, and memory waste.

• Supports various log matching rule writing and abnormal threshold adjustment, and can be optimized according to actual scenarios.

Compass has supported the concept of diagnostic types:

Diagnostic Dimensions	Diagnostic Type	Type Description
Failure analysis	Run failure	Tasks that ultimately fail to run
	First failure	Tasks that have been retried more than once
	Long term failure	Tasks that have failed to run in the last ten days
Time analysis	Baseline time abnormality	Tasks that end earlier or later than the historical normal end time
	Baseline time-consuming abnormality	Tasks that run for too long or too short relative to the historical normal running time
	Long running time	Tasks that run for more than two hours
Error analysis	SQL failure	Tasks that fail due to SQL execution issues
	Shuffle failure	Tasks that fail due to shuffle execution issues
	Memory overflow	Tasks that fail due to memory overflow issues
Cost analysis	Memory waste	Tasks with a peak memory usage to total memory ratio that is too low
	CPU waste	Tasks with a driver/executor calculation time to total CPU calculation time ratio that is too low
Efficiency analysis	Large table scanning	Tasks with too many scanned rows due to no partition restrictions
	OOM warning	Tasks with a cumulative memory of broadcast tables and a high memory ratio of driver or executor
	Data skew	Tasks where the maximum amount of data processed by the task in the stage is much larger than the median
	Job time-consuming abnormality	Tasks with a high ratio of idle time to job running time
	Stage time-consuming abnormality	Tasks with a high ratio of idle time to stage running time
	Task long tail	Tasks where the maximum running time of the task in the stage is much larger than the median
	HDFS stuck	Tasks where the processing rate of tasks in the stage is too slow
	Too many speculative execution tasks	Tasks in which speculative execution of tasks frequently occurs in the stage
	Global sorting abnormality	Tasks with long running time due to global sorting

Get Started

1. Compile

git clone https://github.com/cubefs/compass.git
cd compass
mvn package -DskipTests

2. Configure

cd dist/compass

vi bin/compass_env.sh
# Scheduler MySQL
export SCHEDULER_MYSQL_ADDRESS="ip:port"
export SCHEDULER_MYSQL_DB="scheduler"
export SCHEDULER_DATASOURCE_USERNAME="user"
export SCHEDULER_DATASOURCE_PASSWORD="pwd"
# Compass MySQL
export COMPASS_MYSQL_ADDRESS="ip:port"
export COMPASS_MYSQL_DB="compass"
export SPRING_DATASOURCE_USERNAME="user"
export SPRING_DATASOURCE_PASSWORD="pwd"
# Kafka
export SPRING_KAFKA_BOOTSTRAPSERVERS="ip1:port,ip2:port"
# Redis
export SPRING_REDIS_CLUSTER_NODES="ip1:port,ip2:port"
# Zookeeper
export SPRING_ZOOKEEPER_NODES="ip1:port,ip2:port"
# Elasticsearch
export SPRING_ELASTICSEARCH_NODES="ip1:port,ip2:port"

vi conf/application-hadoop.yml
hadoop:
  namenodes:
    - nameservices: logs-hdfs # the value of dfs.nameservices
      namenodesAddr: [ "machine1.example.com", "machine2.example.com" ] # the value of dfs.namenode.rpc-address.[nameservice ID].[name node ID]
      namenodes: [ "nn1", "nn2" ] # the value of dfs.ha.namenodes.[nameservice ID]
      user: hdfs
      password:
      port: 8020
      # scheduler platform hdfs log path keyword identification, used by task-application
      matchPathKeys: [ "flume" ]

  yarn:
    - clusterName: "bigdata"
      resourceManager: [ "machine1:8088", "machine2:8088" ] # the value of yarn.resourcemanager.webapp.address
      jobHistoryServer: "machine3:19888" # the value of mapreduce.jobhistory.webapp.address

  spark:
    sparkHistoryServer: [ "machine4:18080" ] # the value of spark.history.ui

3. Initialize the database and tables

The Compass table structure consists of two parts, one is compass.sql, and the other is a table that depends on the scheduling platform (dolphinscheduler.sql or airflow.sql, etc.)

1. Please execute document/sql/compass.sql first

2. If you are using the DolphinScheduler scheduling platform, please execute document/sql/dolphinscheduler.sql; if you are using the Airflow scheduling platform, please execute document/sql/airflow.sql

3. If you are using a self-developed scheduling platform, please refer to the task-syncer module to determine the tables that need to be synchronized

4. Deploy

./bin/start_all.sh

Documents

architecture document

deployment document

User Interface

Community

Welcome to join the community for the usage or development of Compass. Here is the way to get help:

• Submit an issue.
• Join the wechat group, search and add WeChat ID daiwei_cn or zebozhuang. Please indicate your intention in the verification information. After verification, we will invite you to the community group.

License

Compass is licensed under the Apache License, Version 2.0 For detail see LICENSE and NOTICE.