kakfa在zookeeper中的节点结构

时间：2022-12-31 13:46:48
kakfa在zookeeper中的节点结构
  • topic注册信息

      /brokers/topics/[topic] : 存储某个topic的partitions所有分配信息

  
  
    

  1. Schema:
    2. 
   
   
    3. 
   
   
  3. { 
    4. 
   
   
  4.  "version": "版本编号目前固定为数字1", 
    5. 
   
   
  5.  "partitions": { "partitionId编号": [ 同步副本组brokerId列表 ], "partitionId编号": [ 同步副本组brokerId列表 ], ....... }}
    6. 
   
   
    7. 
   
   
  7. Example:
    8. 
   
   
  8. {
    9. 
   
   
  9.  "version": 1,
    10. 
   
   
  10.  "partitions": {"0": [1, 2],"1": [2, 1],"2": [1, 2],}
    11. 
   
   
  11. }
    12. 
   
   
  12. 说明：紫红色为patitions编号，蓝色为同步副本组brokerId列表
    13.
  • partition状态信息

   /brokers/topics/[topic]/partitions/[0...N]  其中[0..N]表示partition索引号

  /brokers/topics/[topic]/partitions/[partitionId]/state

  
  
    

  1. Schema:
    2. 
   
   
    3. 
   
   
  3. {"controller_epoch": 表示kafka集群中的*控制器选举次数,"leader": 表示该partition选举leader的brokerId,"version": 版本编号默认为1,"leader_epoch": 该partition leader选举次数,"isr": [同步副本组brokerId列表]}
    4. 
   
   
    5. 
   
   
  5.  
    6. 
   
   
  6. Example:
    7. 
   
   
    8. 
   
   
  8. {
    9. 
   
   
  9.  "controller_epoch": 1,
    10. 
   
   
  10.  "leader": 2,
    11. 
   
   
  11.  "version": 1,
    12. 
   
   
  12.  "leader_epoch": 0,
    13. 
   
   
  13.  "isr": [2, 1]
    14. 
   
   
  14. }
    15.

  • Broker注册信息

   /brokers/ids/[0...N]                 

  每个broker的配置文件中都需要指定一个数字类型的id(全局不可重复),此节点为临时znode(EPHEMERAL)

  
  
    

  1. Schema:
    2. 
   
   
    3. 
   
   
  3. {
    4. 
   
   
  4. "jmx_port": jmx端口号,
    5. 
   
   
  5. "timestamp": kafka broker初始启动时的时间戳,
    6. 
   
   
  6. "host": 主机名或ip地址,
    7. 
   
   
  7. "version": 版本编号默认为1,
    8. 
   
   
  8. "port": kafka broker的服务端端口号,由server.properties中参数port确定
    9. 
   
   
  9. }
    10. 
   
   
  10.  
    11. 
   
   
  11. Example:
    12. 
   
   
    13. 
   
   
  13. {
    14. 
   
   
  14.  "jmx_port": 6061,
    15. 
   
   
  15.  "timestamp":"1403061899859"
    16. 
   
   
  16.  "version": 1,
    17. 
   
   
  17.  "host": "192.168.1.148",
    18. 
   
   
  18.  "port": 9092
    19. 
   
   
  19. }
    20.

  •  Controller epoch

   /controller_epoch -> int (epoch)   

此值为一个数字,kafka集群中第一个broker第一次启动时为1，以后只要集群中center controller*控制器所在broker变更或挂掉，就会重新选举新的center controller，每次center controller变更controller_epoch值就会 + 1; 

  •  Controller注册信息

/controller -> int (broker id of the controller)  存储center controller*控制器所在kafka broker的信息

  
  
    

  1. {
    2. 
   
   
  2. "version": 版本编号默认为1,"brokerid": kafka集群中broker唯一编号,"timestamp": kafka broker*控制器变更时的时间戳
    3. 
   
   
  3. }
    4. 
   
   
    5. 
   
   
  5.  
    6. 
   
   
  6. Example:
    7. 
   
   
    8. 
   
   
    9. 
   
   
  9. {
    10. 
   
   
  10.  "version": 1,
    11. 
   
   
  11.  "brokerid": 3,
    12. 
   
   
  12.  "timestamp": "1403061802981"
    13. 
   
   
  13. }
    14.

  • Consumer注册信息

每个consumer都有一个唯一的ID(consumerId可以通过配置文件指定,也可以由系统生成),此id用来标记消费者信息.

/consumers/[groupId]/ids/[consumerIdString]

是一个临时的znode,此节点的值为请看consumerIdString产生规则,即表示此consumer目前所消费的topic + partitions列表.

consumerId产生规则：

  
  
    

  1.  StringconsumerUuid = null; 
    2. 
   
   
  2.  if(config.consumerId!=null && config.consumerId) 
    3. 
   
   
  3.  consumerUuid = consumerId; 
    4. 
   
   
  4. else 
    5. 
   
   
  5. { 
    6. 
   
   
  6.  String uuid = UUID.randomUUID() 
    7. 
   
   
  7.  consumerUuid = "%s-%d-%s".format(InetAddress.getLocalHost.getHostName, System.currentTimeMillis, uuid.getMostSignificantBits().toHexString.substring(0,8)); 
    8. 
   
   
  8. } 
    9. 
   
   
  9. String consumerIdString = config.groupId + "_" + consumerUuid;
    10. 
  
  
   
   
    

  1. Schema:
    2. 
    
    
    3. 
    
    
  3. {"version": 版本编号默认为1,
    4. 
    
    
  4. "subscription": { //订阅topic列表
    5. 
    
    
  5. "topic名称": consumer中topic消费者线程数
    6. 
    
    
  6. },
    7. 
    
    
  7. "pattern": "static",
    8. 
    
    
  8. "timestamp": "consumer启动时的时间戳"
    9. 
    
    
  9. }
    10. 
    
    
    11. 
    
    
  11.  
    12. 
    
    
  12. Example:
    13. 
    
    
  13. {
    14. 
    
    
  14. "version": 1,
    15. 
    
    
  15. "subscription": {"open_platform_opt_push_plus1": 5},
    16. 
    
    
  16. "pattern": "static",
    17. 
    
    
  17. "timestamp": "1411294187842"
    18. 
    
    
  18. } 
    19.

  • Consumer owner

   /consumers/[groupId]/owners/[topic]/[partitionId] -> consumerIdString + threadId索引编号

   当consumer启动时,所触发的操作:

    • 首先进行"Consumer Id注册";
    • 然后在"Consumer id 注册"节点下注册一个watch用来监听当前group中其他consumer的"退出"和"加入";只要此znode path下节点列表变更,都会触发此group下consumer的负载均衡.(比如一个consumer失效,那么其他consumer接管partitions)
    • 在"Broker id 注册"节点下,注册一个watch用来监听broker的存活情况;如果broker列表变更,将会触发所有的groups下的consumer重新balance.
  • Consumer offset

   /consumers/[groupId]/offsets/[topic]/[partitionId] -> long (offset)

  用来跟踪每个consumer目前所消费的partition中最大的offset

  此znode为持久节点,可以看出offset跟group_id有关,以表明当消费者组(consumer group)中一个消费者失效,  重新触发   balance,其他consumer可以继续消费.

  • Re-assign partitions

  /admin/reassign_partitions

  
  
    

  1. {
    2. 
   
   
  2.  "fields":[
    3. 
   
   
  3.  {
    4. 
   
   
  4.  "name":"version",
    5. 
   
   
  5.  "type":"int",
    6. 
   
   
  6.  "doc":"version id"
    7. 
   
   
  7.  },
    8. 
   
   
  8.  {
    9. 
   
   
  9.  "name":"partitions",
    10. 
   
   
  10.  "type":{
    11. 
   
   
  11.  "type":"array",
    12. 
   
   
  12.  "items":{
    13. 
   
   
  13.  "fields":[
    14. 
   
   
  14.  {
    15. 
   
   
  15.  "name":"topic",
    16. 
   
   
  16.  "type":"string",
    17. 
   
   
  17.  "doc":"topic of the partition to be reassigned"
    18. 
   
   
  18.  },
    19. 
   
   
  19.  {
    20. 
   
   
  20.  "name":"partition",
    21. 
   
   
  21.  "type":"int",
    22. 
   
   
  22.  "doc":"the partition to be reassigned"
    23. 
   
   
  23.  },
    24. 
   
   
  24.  {
    25. 
   
   
  25.  "name":"replicas",
    26. 
   
   
  26.  "type":"array",
    27. 
   
   
  27.  "items":"int",
    28. 
   
   
  28.  "doc":"a list of replica ids"
    29. 
   
   
  29.  }
    30. 
   
   
  30.  ],
    31. 
   
   
  31.  }
    32. 
   
   
  32.  "doc":"an array of partitions to be reassigned to new replicas"
    33. 
   
   
  33.  }
    34. 
   
   
  34.  }
    35. 
   
   
  35.  ]
    36. 
   
   
  36. }
    37. 
   
   
  37.  
    38. 
   
   
  38. Example:
    39. 
   
   
  39. {
    40. 
   
   
  40.  "version": 1,
    41. 
   
   
  41.  "partitions":
    42. 
   
   
  42.  [
    43. 
   
   
  43.  {
    44. 
   
   
  44.  "topic": "Foo",
    45. 
   
   
  45.  "partition": 1,
    46. 
   
   
  46.  "replicas": [0, 1, 3]
    47. 
   
   
  47.  }
    48. 
   
   
  48.  ] 
    49. 
   
   
  49. }
    50.

  •  Preferred replication election

   /admin/preferred_replica_election

 

  
  
    

  1. {
    2. 
   
   
  2.  "fields":[
    3. 
   
   
  3.  {
    4. 
   
   
  4.  "name":"version",
    5. 
   
   
  5.  "type":"int",
    6. 
   
   
  6.  "doc":"version id"
    7. 
   
   
  7.  },
    8. 
   
   
  8.  {
    9. 
   
   
  9.  "name":"partitions",
    10. 
   
   
  10.  "type":{
    11. 
   
   
  11.  "type":"array",
    12. 
   
   
  12.  "items":{
    13. 
   
   
  13.  "fields":[
    14. 
   
   
  14.  {
    15. 
   
   
  15.  "name":"topic",
    16. 
   
   
  16.  "type":"string",
    17. 
   
   
  17.  "doc":"topic of the partition for which preferred replica election should be triggered"
    18. 
   
   
  18.  },
    19. 
   
   
  19.  {
    20. 
   
   
  20.  "name":"partition",
    21. 
   
   
  21.  "type":"int",
    22. 
   
   
  22.  "doc":"the partition for which preferred replica election should be triggered"
    23. 
   
   
  23.  }
    24. 
   
   
  24.  ],
    25. 
   
   
  25.  }
    26. 
   
   
  26.  "doc":"an array of partitions for which preferred replica election should be triggered"
    27. 
   
   
  27.  }
    28. 
   
   
  28.  }
    29. 
   
   
  29.  ]
    30. 
   
   
  30. }
    31. 
   
   
  31.  
    32. 
   
   
  32. 例子:
    33. 
   
   
  33.  
    34. 
   
   
  34. {
    35. 
   
   
  35.  "version": 1,
    36. 
   
   
  36.  "partitions":
    37. 
   
   
  37.  [
    38. 
   
   
  38.  {
    39. 
   
   
  39.  "topic": "Foo",
    40. 
   
   
  40.  "partition": 1 
    41. 
   
   
  41.  },
    42. 
   
   
  42.  {
    43. 
   
   
  43.  "topic": "Bar",
    44. 
   
   
  44.  "partition": 0 
    45. 
   
   
  45.  }
    46. 
   
   
  46.  ] 
    47. 
   
   
  47. }
    48.

  • 删除topics

  /admin/delete_topics

  
  
    

  1. Schema:
    2. 
   
   
  2. { "fields":
    3. 
   
   
  3.  [ {"name": "version", "type": "int", "doc": "version id"},
    4. 
   
   
  4.  {"name": "topics",
    5. 
   
   
  5.  "type": { "type": "array", "items": "string", "doc": "an array of topics to be deleted"}
    6. 
   
   
  6.  } ]
    7. 
   
   
  7. }
    8. 
   
   
  8.  
    9. 
   
   
  9. 例子:
    10. 
   
   
  10. {
    11. 
   
   
  11.  "version": 1,
    12. 
   
   
  12.  "topics": ["foo", "bar"]
    13. 
   
   
  13. }
    14.

  • Topic配置

  /config/topics/[topic_name]

  
  
    

  1. {
    2. 
   
   
  2.  "version": 1,
    3. 
   
   
  3.  "config": {
    4. 
   
   
  4.  "config.a": "x",
    5. 
   
   
  5.  "config.b": "y",
    6. 
   
   
  6.  ...
    7. 
   
   
  7.  }
    8. 
   
   
  8. }
    9.
标签：

相关文章