mplachter.flume

Apache-Flume =========

用于部署和配置 Apache Flume 的 Ansible 角色

• 部署 Apache Flume
• 配置 Apache Flume
• 创建 apache-flume 服务
  • 仅运行 agent 配置

要求

• 运行
  • Ansible 2.3+
• 测试
  • Docker/Vagrant
  • Molecule = 1.2.5

角色变量

• Java vars

  java_heap_xms: 125
  java_heap_xmx: 250
  

• Apache Flume vars

  mirror_url: http://apache.mirrors.ionfish.org/flume
  version: 1.7.0
  

• Linux 文件夹/路径安装 vars

  download_path: /tmp
  installation_path: /usr/local
  owner: root
  group: root
  

• Apache Flume 配置 vars
  • 请参考 Flume 用户指南
  • 当前配置将允许
    • 代理
      • 服务将仅为 agent 创建
      • 多个
      • 当前需要手动创建更多服务来运行这些
    • 源
      • 当前每个 agent 仅允许一个源
    • 通道
      • 当前每个 agent 仅允许一个通道
    • 汇
      • 多个
    • 汇组
      • 将所有 sinks 添加到 sinkgroup
  • 由于 Flume 配置选项非常多，请阅读以下内容
    • 请在 Flume 配置的 属性名 中将 "." 替换为 "_"
      • 值不需要替换
      • 示例
        • kafka_consumer_group_id: testflume
      • 结果
        • agent.source.kafka.consumer.group.id = testflume
  • 可以传递 apache_flume_config 变量以复制配置
    • 示例
      • apache_flume_config: file/flume-conf.properties
    • 结果
      • 这将把 flume-conf.properties 从你的文件目录复制到目标机器
• 额外变量
  • HDFS 原生库
    • hdfs_libs = true
      • 这将下载 HDFS 原生库 并放置在 plugin.d/hdfs/native/
• 示例变量

  mirror_url: http://apache.mirrors.ionfish.org/flume
  version: 1.7.0
  download_path: /tmp
  installation_path: /usr/local
  owner: root
  group: root
  
  java_heap_xms: 125
  java_heap_xmx: 250
  hdfs_libs: true
  
  agents:
  - name: agent
    source:
      name: kafkaSource
      type: org.apache.flume.source.kafka.KafkaSource
      kafka_consumer_group_id: flume
      kafka_consumer_auto_offset_reset: latest
      kafka_consumer_max_partition_fetch_bytes: 1048576
      kafka_consumer_heartbeat_interval_ms: 3000
      kafka_consumer_session_timeout_ms: 30000
      kafka_consumer_request_timeout_ms: 40000
      kafka_consumer_fetch_max_wait_ms: 500
      kafka_bootstrap_servers:
        - 127.0.0.1:9092
        - 0.0.0.0:9092
      kafka_topics:
        - topic1
        - topic2
    channel:
      name: kakfaChannel
      type: memory
      capacity: 1000000
      transactionCapacity: 100000
    sinks:
      - name: kafkaHDFSSink1
        type: hdfs
        hdfs_path: "s3n://GFGJFSHFJHFGFHSBJ:fdjhSFUYGSF65678+-saigfew123@hdfs/%{topic}/%y/%m/%d/%H"
        hdfs_filePrefix: FlumeData
        hdfs_inUseSuffix: .tmp
        hdfs_rollInterval: 30
        hdfs_rollSize: 1024
        hdfs_rollCount: 10
        hdfs_idleTimeout: 0
        hdfs_batchSize: 100
        hdfs_fileType: "SequenceFile"
        hdfs_maxOpenFiles: 5000
        hdfs_callTimeout: 10000
        hdfs_threadsPoolSize: 10
        hdfs_rollTimerPoolSize: 1
        hdfs_round: false
        hdfs_roundValue: 1
        hdfs_roundUnit: second
        hdfs_timeZone: Local Time
        hdfs_useLocalTimeStamp: false
        hdfs_closeTries: 0
        hdfs_retryInterval: 180
      - name: kafkaHDFSSink2
        type: hdfs
        hdfs_path: "s3n://GFGJFSHFJHFGFHSBJ:fdjhSFUYGSF65678+-saigfew123@hdfs/%{topic}/%y/%m/%d/%H"
        hdfs_filePrefix: FlumeData
        hdfs_inUseSuffix: .tmp
        hdfs_rollInterval: 30
        hdfs_rollSize: 1024
        hdfs_rollCount: 10
        hdfs_idleTimeout: 0
        hdfs_batchSize: 100
        hdfs_fileType: "SequenceFile"
        hdfs_maxOpenFiles: 5000
        hdfs_callTimeout: 10000
        hdfs_threadsPoolSize: 10
        hdfs_rollTimerPoolSize: 1
        hdfs_round: false
        hdfs_roundValue: 1
        hdfs_roundUnit: second
        hdfs_timeZone: Local Time
        hdfs_useLocalTimeStamp: false
        hdfs_closeTries: 0
        hdfs_retryInterval: 180
    sink_group:
      name: sinkgroup1
      processor_type: load_balance
      processor_backoff: false
      processor_selector: round_robin
  

依赖

• andrewrothstein.java-oracle-jre

示例剧本

- hosts: all
  roles:
    - role: mplachter.flume

许可证

MIT

作者信息

Matthew Plachter