Service.properties arguments detailed
server.properties is Kafka's main configuration file, the following brief meaning of which configuration items. The core of the three most general configuration broker.id, log.dir, zookeeper.connect. Other configurations can be modified according to their own needs.
Enter the configuration directory, usually associated configuration mainly in the following three
Other configuration is as follows:
| Service.properties arguments detailed |
||||
| parameter |
Functional Description |
Recommendations |
||
| broker.id=0 |
Each Broker-only identity in the cluster. Broker even if the IP address has changed, broker.id as long as no change, then the message will not affect the situation of consumers |
Nowhere to be modified to a different broker Id, is typically an integer |
||
| log.dirs=/data/kafka-logs |
Kafka storage address data, the plurality of addresses if a comma-separated, a plurality of directories are distributed on different disk read and write performance can be increased / data / kafka-logs-1, / data / kafka-logs-2 |
It needs to be modified according to the directory using the general |
||
| port =9092 |
broker server service port |
|
||
| message.max.bytes =6525000 |
It represents the maximum size of the message body, in bytes |
|
||
| num.network.threads =4 |
broker maximum number of threads to process messages, in general, the number is the number of cpu core |
|
||
| num.io.threads =8 |
broker number of processing threads disk IO, a value of two times the number of cpu core |
|
||
| background.threads =4 |
Some background task processing threads, for example, delete outdated message file, under normal circumstances you do not need to modify |
|
||
| queued.max.requests =500 |
IO Threading waiting queue maximum number of requests, if the request is waiting for IO exceeds this value, it will stop accepting external message, it should be a self-protection mechanism. |
|
||
| host.name |
broker host address, if set up, it will bind to this address, if not, it will bind to all interfaces, and send one of them to ZK, generally do not set |
|
||
| socket.send.buffer.bytes=100*1024 |
The socket transmit buffer, socket tuning parameters SO_SNDBUFF |
|
||
| socket.receive.buffer.bytes =100*1024 |
the socket receive buffer, socket tuning parameters SO_RCVBUFF |
|
||
| socket.request.max.bytes =100*1024*1024 |
socket requested maximum value, when the cover to prevent the specified parameters serverOOM, message.max.bytes bound to less than socket.request.max.bytes, will be created topic |
|
||
| log.segment.bytes =1024*1024*1024 |
topic partition based on a bunch of files stored in the segment, the control of the size of each segment, covering the specified parameters will be created when the topic |
Log segment size |
||
| log.roll.hours =24*7 |
This parameter does not reach the size of log.segment.bytes set in the log segment, covering the time specified parameters will be forced to create a new segment will be created topic |
|
||
| log.cleanup.policy = delete |
Log cleanup strategy choose: delete expired and compact mainly for data processing, or log file reaches the limit on the amount of specified parameters will be created when the topic covered |
|
||
| log.retention.minutes=300 |
How long to keep data files, maximum storage time than this time will clear strategy based on log.cleanup.policy setting data |
|
||
| log.retention.bytes=-1 |
topic每个分区的最大文件大小，一个topic的大小限制 = 分区数*log.retention.bytes。-1没有大小限log.retention.bytes和log.retention.minutes任意一个达到要求，都会执行删除，会被topic创建时的指定参数覆盖 |
必须限制日志文件的最大大小，防止出现日志占满磁盘的情况 |
||
| log.retention.check.interval.ms=5minutes |
文件大小检查的周期时间，是否处罚 log.cleanup.policy中设置的策略 |
|
||
| log.cleaner.enable=false |
是否开启日志清理 |
|
||
| log.cleaner.threads = 2 |
日志清理运行的线程数 |
|
||
| log.cleaner.io.max.bytes.per.second=None |
日志清理时候处理的最大大小 |
|
||
| log.cleaner.dedupe.buffer.size=500*1024*1024 |
日志清理去重时候的缓存空间，在空间允许的情况下，越大越好 |
|
||
| log.cleaner.io.buffer.size=512*1024 |
日志清理时候用到的IO块大小一般不需要修改 |
|
||
| log.cleaner.io.buffer.load.factor =0.9 |
日志清理中hash表的扩大因子一般不需要修改 |
|
||
| log.cleaner.backoff.ms =15000 |
检查是否处罚日志清理的间隔 |
|
||
| log.cleaner.min.cleanable.ratio=0.5 |
日志清理的频率控制，越大意味着更高效的清理，同时会存在一些空间上的浪费，会被topic创建时的指定参数覆盖 |
|
||
| log.cleaner.delete.retention.ms =1day |
对于压缩的日志保留的最长时间，也是客户端消费消息的最长时间，同log.retention.minutes的区别在于一个控制未压缩数据，一个控制压缩后的数据。会被topic创建时的指定参数覆盖 |
|
||
| log.index.size.max.bytes =10*1024*1024 |
对于segment日志的索引文件大小限制，会被topic创建时的指定参数覆盖 |
|
||
| log.index.interval.bytes =4096 |
当执行一个fetch操作后，需要一定的空间来扫描最近的offset大小，设置越大，代表扫描速度越快，但是也更好内存，一般情况下不需要搭理这个参数 |
|
||
| log.flush.interval.messages=None |
log文件”sync”到磁盘之前累积的消息条数,因为磁盘IO操作是一个慢操作,但又是一个”数据可靠性"的必要手段,所以此参数的设置,需要在"数据可靠性"与"性能"之间做必要的权衡.如果此值过大,将会导致每次"fsync"的时间较长(IO阻塞),如果此值过小,将会导致"fsync"的次数较多,这也意味着整体的client请求有一定的延迟.物理server故障,将会导致没有fsync的消息丢失. |
|
||
| log.flush.scheduler.interval.ms =3000 |
检查是否需要固化到硬盘的时间间隔 |
|
||
| log.flush.interval.ms = None |
仅仅通过interval来控制消息的磁盘写入时机,是不足的.此参数用于控制"fsync"的时间间隔,如果消息量始终没有达到阀值,但是离上一次磁盘同步的时间间隔达到阀值,也将触发. |
|
||
| log.delete.delay.ms =60000 |
文件在索引中清除后保留的时间一般不需要去修改 |
|
||
| log.flush.offset.checkpoint.interval.ms =60000 |
控制上次固化硬盘的时间点，以便于数据恢复一般不需要去修改 |
|
||
| auto.create.topics.enable =true |
是否允许自动创建topic，若是false，就需要通过命令创建topic |
|
||
| default.replication.factor =1 |
一个topic ，默认分区的replication个数 ，不得大于集群中broker的个数 |
|
||
| num.partitions =1 |
每个topic的分区个数，若是在topic创建时候没有指定的话会被topic创建时的指定参数覆盖 |
|
||
| controller.socket.timeout.ms =30000 |
partition leader与replicas之间通讯时,socket的超时时间 |
|
||
| controller.message.queue.size=10 |
partition leader与replicas数据同步时,消息的队列尺寸 |
|
||
| replica.lag.time.max.ms =10000 |
replicas响应partition leader的最长等待时间，若是超过这个时间，就将replicas列入ISR(in-sync replicas)，并认为它是死的，不会再加入管理中 |
|
||
| replica.lag.max.messages =4000 |
如果follower落后与leader太多,将会认为此follower[或者说partition relicas]已经失效 |
|
||
| replica.socket.timeout.ms=30*1000 |
follower与leader之间的socket超时时间 |
|
||
| replica.socket.receive.buffer.bytes=64*1024 |
leader复制时候的socket缓存大小 |
|
||
| replica.fetch.max.bytes =1024*1024 |
replicas每次获取数据的最大大小 |
|
||
| replica.fetch.wait.max.ms =500 |
replicas同leader之间通信的最大等待时间，失败了会重试 |
|
||
| replica.fetch.min.bytes =1 |
fetch的最小数据尺寸,如果leader中尚未同步的数据不足此值,将会阻塞,直到满足条件 |
|
||
| num.replica.fetchers=1 |
leader进行复制的线程数，增大这个数值会增加follower的IO |
|
||
| replica.high.watermark.checkpoint.interval.ms =5000 |
每个replica检查是否将最高水位进行固化的频率 |
|
||
| controlled.shutdown.enable =false |
是否允许控制器关闭broker ,若是设置为true,会关闭所有在这个broker上的leader，并转移到其他broker |
|
||
| controlled.shutdown.max.retries =3 |
控制器关闭的尝试次数 |
|
||
| controlled.shutdown.retry.backoff.ms =5000 |
每次关闭尝试的时间间隔 |
|
||
| leader.imbalance.per.broker.percentage =10 |
leader的不平衡比例，若是超过这个数值，会对分区进行重新的平衡 |
|
||
| leader.imbalance.check.interval.seconds =300 |
检查leader是否不平衡的时间间隔 |
|
||
| offset.metadata.max.bytes |
客户端保留offset信息的最大空间大小 |
|
||
| zookeeper.connect = localhost:2181 |
zookeeper集群的地址，可以是多个，多个之间用逗号分割hostname1:port1,hostname2:port2,hostname3:port3 |
|
||
| zookeeper.session.timeout.ms=6000 |
ZooKeeper的最大超时时间，就是心跳的间隔，若是没有反映，那么认为已经死了，不易过大 |
|
||
| zookeeper.connection.timeout.ms =6000 |
ZooKeeper的连接超时时间 |
|
||
| zookeeper.sync.time.ms =2000 |
ZooKeeper集群中leader和follower之间的同步时间 |
|
||
| consumer.properties参数详解 |
||||
| 参数 |
功能描述 |
使用建议 |
||
| group.id |
Consumer归属的组ID，broker是根据group.id来判断是队列模式还是发布订阅模式，非常重要 |
|
||
| consumer.id |
消费者的ID，若是没有设置的话，会自增 |
|
||
| client.id = group id value |
一个用于跟踪调查的ID ，最好同group.id相同 |
|
||
| zookeeper.connect=localhost:2182 |
对于zookeeper集群的指定，可以是多个 hostname1:port1,hostname2:port2,hostname3:port3 必须和broker使用同样的zk配置 |
|
||
| zookeeper.session.timeout.ms = 6000 |
zookeeper的心跳超时时间，查过这个时间就认为是dead消费者 |
|
||
| zookeeper.connection.timeout.ms = 6000 |
zookeeper的等待连接时间 |
|
||
| zookeeper.sync.time.ms = 2000 |
zookeeper的follower同leader的同步时间 |
|
||
| auto.offset.reset = largest |
当zookeeper中没有初始的offset时候的处理方式 。smallest ：重置为最小值 largest:重置为最大值 anything else：抛出异常 |
|
||
| socket.timeout.ms= 30 * 1000 |
socket的超时时间，实际的超时时间是：max.fetch.wait + socket.timeout.ms. |
|
||
| socket.receive.buffer.bytes=64 * 1024 |
socket的接受缓存空间大小 |
|
||
| fetch.message.max.bytes = 1024 * 1024 |
从每个分区获取的消息大小限制 |
|
||
| auto.commit.interval.ms = 60 * 1000 |
自动提交的时间间隔 |
|
||
| queued.max.message.chunks = 10 |
用来处理消费消息的块，每个块可以等同于fetch.message.max.bytes中数值 |
|
||
| rebalance.max.retries = 4 |
## 当有新的consumer加入到group时,将会reblance,此后将会有partitions的消费端迁移到新的consumer上,如果一个consumer获得了某个partition的消费权限,那么它将会向zk注册"Partition Owner registry"节点信息,但是有可能此时旧的consumer尚没有释放此节点, |
|
||
| rebalance.backoff.ms = 2000 |
每次再平衡的时间间隔 |
|
||
| refresh.leader.backoff.ms |
每次重新选举leader的时间 |
|
||
| fetch.min.bytes = 1 |
server发送到消费端的最小数据，若是不满足这个数值则会等待，知道满足数值要求 |
|
||
| fetch.wait.max.ms = 100 |
若是不满足最小大小(fetch.min.bytes)的话，等待消费端请求的最长等待时间 |
|
||
| consumer.timeout.ms = -1 |
指定时间内没有消息到达就抛出异常，一般不需要改 |
|
||
| producer.properties参数详解 |
||
| 参数 |
功能描述 |
使用建议 |
| metadata.broker.list |
消费者获取消息元信息(topics, partitions and replicas)的地址,配置格式是：host1:port1,host2:port2，也可以在外面设置一个vip |
|
| request.required.acks = 0 |
##消息的确认模式 |
|
| request.timeout.ms = 10000 |
消息发送的最长等待时间 |
|
| send.buffer.bytes=100*1024 |
socket的缓存大小 |
|
| key.serializer.class |
key的序列化方式，若是没有设置，同serializer.class |
|
| partitioner.class=kafka.producer.DefaultPartitioner |
分区的策略，默认是取模 |
|
| compression.codec = none |
消息的压缩模式，默认是none，可以有gzip和snappy |
|
| compressed.topics=null |
可以针对默写特定的topic进行压缩 |
|
| message.send.max.retries = 3 |
消息发送失败后的重试次数 |
|
| retry.backoff.ms = 100 |
每次失败后的间隔时间 |
|
| topic.metadata.refresh.interval.ms = 600 * 1000 |
生产者定时更新topic元信息的时间间隔 ，若是设置为0，那么会在每个消息发送后都去更新数据 |
|
| client.id="" |
用户随意指定，但是不能重复，主要用于跟踪记录消息 |
|
| producer.type=sync |
生产者的类型 async:异步执行消息的发送 sync：同步执行消息的发送 |
|
| queue.buffering.max.ms = 5000 |
异步模式下，那么就会在设置的时间缓存消息，并一次性发送 |
|
| queue.buffering.max.messages = 10000 |
异步的模式下 最长等待的消息数 |
|
| queue.enqueue.timeout.ms = -1 |
异步模式下，进入队列的等待时间 若是设置为0，那么要么进入队列，要么直接抛弃 |
|
| batch.num.messages=200 |
异步模式下，每次发送的最大消息数，前提是触发了queue.buffering.max.messages或是queue.buffering.max.ms的限制 |
|
| serializer.class = kafka.serializer.DefaultEncoder |
消息体的系列化处理类 ，转化为字节流进行传输 |
|
参考：
https://blog.csdn.net/selfsojourner/article/details/44957135
https://blog.csdn.net/lizhitao/article/details/25667831