http://blog.javachen.com/2015/03/17/install-and-test-kafka/#
本文主要介紹如何在單節點上安裝 Kafka 並測試 broker、producer 和 consumer 功能。
下載
進入下載頁面:http://kafka.apache.org/downloads.html ,選擇 Binary downloads下載 (Source download需要編譯才能使用),這裏我下載 kafka_2.11-0.8.2.1
,其對應的
Scala 版本爲 2.11
:
$ wget http://apache.fayea.com/kafka/0.8.2.1/kafka_2.11-0.8.2.1.tgz
解壓並進入目錄:
$ tar -xzvf kafka_2.11-0.8.2.1.tgz
$ cd kafka_2.11-0.8.2.1
查看目錄結構:
tree -L 2
.
├── bin
│ ├── kafka-console-consumer.sh
│ ├── kafka-console-producer.sh
│ ├── kafka-consumer-offset-checker.sh
│ ├── kafka-consumer-perf-test.sh
│ ├── kafka-mirror-maker.sh
│ ├── kafka-preferred-replica-election.sh
│ ├── kafka-producer-perf-test.sh
│ ├── kafka-reassign-partitions.sh
│ ├── kafka-replay-log-producer.sh
│ ├── kafka-replica-verification.sh
│ ├── kafka-run-class.sh
│ ├── kafka-server-start.sh
│ ├── kafka-server-stop.sh
│ ├── kafka-simple-consumer-shell.sh
│ ├── kafka-topics.sh
│ ├── windows
│ ├── zookeeper-server-start.sh
│ ├── zookeeper-server-stop.sh
│ └── zookeeper-shell.sh
├── config
│ ├── consumer.properties
│ ├── log4j.properties
│ ├── producer.properties
│ ├── server.properties
│ ├── test-log4j.properties
│ ├── tools-log4j.properties
│ └── zookeeper.properties
├── libs
│ ├── jopt-simple-3.2.jar
│ ├── kafka_2.11-0.8.2.1.jar
│ ├── kafka_2.11-0.8.2.1-javadoc.jar
│ ├── kafka_2.11-0.8.2.1-scaladoc.jar
│ ├── kafka_2.11-0.8.2.1-sources.jar
│ ├── kafka_2.11-0.8.2.1-test.jar
│ ├── kafka-clients-0.8.2.1.jar
│ ├── log4j-1.2.16.jar
│ ├── lz4-1.2.0.jar
│ ├── metrics-core-2.2.0.jar
│ ├── scala-library-2.11.5.jar
│ ├── scala-parser-combinators_2.11-1.0.2.jar
│ ├── scala-xml_2.11-1.0.2.jar
│ ├── slf4j-api-1.7.6.jar
│ ├── slf4j-log4j12-1.6.1.jar
│ ├── snappy-java-1.1.1.6.jar
│ ├── zkclient-0.3.jar
│ └── zookeeper-3.4.6.jar
├── LICENSE
└── NOTICE
4 directories, 45 files
啓動和停止
運行 kafka ,需要依賴 zookeeper,你可以使用已有的 zookeeper 集羣或者利用 kafka 提供的腳本啓動一個 zookeeper 實例:
$ bin/zookeeper-server-start.sh config/zookeeper.properties &
默認的,zookeeper 會監聽在 *:2181/tcp
。
停止剛纔啓動的 zookeeper 實例:
$ bin/zookeeper-server-stop.sh
啓動Kafka server:
$ bin/kafka-server-start.sh config/server.properties &
config/server.properties 中有一些默認的配置參數,這裏僅僅列出參數,不做解釋:
broker.id=0
port=9092
#host.name=localhost
#advertised.host.name=<hostname routable by clients>
#advertised.port=<port accessible by clients>
num.network.threads=3
num.io.threads=8
socket.send.buffer.bytes=102400
socket.receive.buffer.bytes=102400
socket.request.max.bytes=104857600
log.dirs=/tmp/kafka-logs
num.partitions=1
num.recovery.threads.per.data.dir=1
#log.flush.interval.messages=10000
#log.flush.interval.ms=1000
log.retention.hours=168
#log.retention.bytes=1073741824
log.segment.bytes=1073741824
log.retention.check.interval.ms=300000
log.cleaner.enable=false
zookeeper.connect=localhost:2181
zookeeper.connection.timeout.ms=6000
如果你像我一樣是在虛擬機中測試 kafka,那麼你需要修改 kafka 啓動參數中 JVM 內存大小。查看 kafka-server-start.sh 腳本,修改 KAFKA_HEAP_OPTS
處 -Xmx
和 -Xms
的值。
啓動成功之後,會看到如下日誌:
[2015-03-17 11:19:30,528] INFO Starting log flusher with a default period of 9223372036854775807 ms. (kafka.log.LogManager)
[2015-03-17 11:19:30,604] INFO Awaiting socket connections on 0.0.0.0:9092. (kafka.network.Acceptor)
[2015-03-17 11:19:30,605] INFO [Socket Server on Broker 0], Started (kafka.network.SocketServer)
[2015-03-17 11:19:30,687] INFO Will not load MX4J, mx4j-tools.jar is not in the classpath (kafka.utils.Mx4jLoader$)
[2015-03-17 11:19:30,756] INFO 0 successfully elected as leader (kafka.server.ZookeeperLeaderElector)
[2015-03-17 11:19:30,887] INFO Registered broker 0 at path /brokers/ids/0 with address cdh1:9092. (kafka.utils.ZkUtils$)
[2015-03-17 11:19:30,928] INFO [Kafka Server 0], started (kafka.server.KafkaServer)
[2015-03-17 11:19:31,048] INFO New leader is 0 (kafka.server.ZookeeperLeaderElector$LeaderChangeListener)
從日誌可以看到:
- log flusher 有一個默認的週期值
- kafka server 監聽在9092端口
- 在 cdh1:9092 上註冊了一個 broker 0 ,路徑爲 /brokers/ids/0
停止 Kafka server :
$ bin/kafka-server-stop.sh
單 broker 測試
在啓動 kafka-server 之後啓動,運行producer:
$ bin/kafka-console-producer.sh --broker-list localhost:9092 --topic test
在另一個終端運行 consumer:
$ bin/kafka-console-consumer.sh --zookeeper localhost:2181 --topic test --from-beginning
在 producer 端輸入字符串並回車,查看 consumer 端是否顯示。
多 broker 測試
配置和啓動 Kafka broker
接下來參考 Running a Multi-Broker Apache Kafka 0.8 Cluster on a Single Node 這篇文章,基於 config/server.properties 配置文件創建多個 broker 的 kafka 集羣。
創建第一個 broker:
$ cp config/server.properties config/server1.properties
編寫 config/server1.properties 並修改下面配置:
broker.id=1
port=9092
log.dir=/tmp/kafka-logs-1
創建第二個 broker:
$ cp config/server.properties config/server2.properties
編寫 config/server2.properties 並修改下面配置:
broker.id=2
port=9093
log.dir=/tmp/kafka-logs-2
創建第三個 broker:
$ cp config/server.properties config/server3.properties
編寫 config/server3.properties 並修改下面配置:
broker.id=3
port=9094
log.dir=/tmp/kafka-logs-3
接下來分別啓動這三個 broker:
$ JMX_PORT=9999 ; nohup bin/kafka-server-start.sh config/server1.properties &
$ JMX_PORT=10000 ; nohup bin/kafka-server-start.sh config/server2.properties &
$ JMX_PORT=10001 ; nohup bin/kafka-server-start.sh config/server3.properties &
下面是三個 broker 監聽的網絡接口和端口列表:
Broker 1 Broker 2 Broker 3
----------------------------------------------
Kafka *:9092/tcp *:9093/tcp *:9094/tcp
JMX *:9999/tcp *:10000/tcp *:10001/tcp
創建 Kafka topic
在 Kafka 0.8 中有兩種方式創建一個新的 topic:
- 在 broker 上開啓
auto.create.topics.enable
參數,當 broker 接收到一個新的 topic 上的消息時候,會通過num.partitions
和default.replication.factor
兩個參數自動創建 topic。 - 使用
bin/kafka-topics.sh
命令
創建一個名稱爲 zerg.hydra
的 topic:
$ bin/kafka-topics.sh --zookeeper localhost:2181 --create --topic zerg.hydra --partitions 3 --replication-factor 2
使用下面查看創建的 topic:
$ bin/kafka-topics.sh --zookeeper localhost:2181 --list
test
zerg.hydra
還可以查看更詳細的信息:
$ bin/kafka-topics.sh --zookeeper localhost:2181 --describe --topic zerg.hydra
Topic:zerg.hydra PartitionCount:3 ReplicationFactor:2 Configs:
Topic: zerg.hydra Partition: 0 Leader: 2 Replicas: 2,3 Isr: 2,3
Topic: zerg.hydra Partition: 1 Leader: 3 Replicas: 3,0 Isr: 3,0
Topic: zerg.hydra Partition: 2 Leader: 0 Replicas: 0,2 Isr: 0,2
默認的,Kafka 持久化 topic 到 log.dir
參數定義的目錄。
$ tree /tmp/kafka-logs-{1,2,3}
/tmp/kafka-logs-1 # first broker (broker.id = 1)
├── zerg.hydra-0 # replica of partition 0 of topic "zerg.hydra" (this broker is leader)
│ ├── 00000000000000000000.index
│ └── 00000000000000000000.log
├── zerg.hydra-2 # replica of partition 2 of topic "zerg.hydra"
│ ├── 00000000000000000000.index
│ └── 00000000000000000000.log
└── replication-offset-checkpoint
/tmp/kafka-logs-2 # second broker (broker.id = 2)
├── zerg.hydra-0 # replica of partition 0 of topic "zerg.hydra"
│ ├── 00000000000000000000.index
│ └── 00000000000000000000.log
├── zerg.hydra-1 # replica of partition 1 of topic "zerg.hydra" (this broker is leader)
│ ├── 00000000000000000000.index
│ └── 00000000000000000000.log
└── replication-offset-checkpoint
/tmp/kafka-logs-3 # third broker (broker.id = 3)
├── zerg.hydra-1 # replica of partition 1 of topic "zerg.hydra"
│ ├── 00000000000000000000.index
│ └── 00000000000000000000.log
├── zerg.hydra-2 # replica of partition 2 of topic "zerg.hydra" (this broker is leader)
│ ├── 00000000000000000000.index
│ └── 00000000000000000000.log
└── replication-offset-checkpoint
6 directories, 15 files
啓動一個 producer
以 sync
模式啓動一個 producer:
$ bin/kafka-console-producer.sh --broker-list localhost:9092,localhost:9093,localhost:9094 --sync --topic zerg.hydra
然後,輸入以下內容:
Hello, world!
Rock: Nerf Paper. Scissors is fine.
啓動一個 consumer
在另一個終端運行:
$ bin/kafka-console-consumer.sh --zookeeper localhost:2181 --topic zerg.hydra --from-beginning
注意,生產環境通常不會添加 --from-beginning
參數。
觀察輸出,你會看到下面內容:
Hello, world!
Rock: Nerf Paper. Scissors is fine.
把 consumer 停掉再啓動,你還會看到相同的輸出結果。
將日誌推送到 kafka
例如,將 apache 或者 nginx 或者 tomcat 等產生的日誌 push 到 kafka,只需要執行下面代碼即可:
$ tail -n 0 -f /var/log/nginx/access.log | bin/kafka-console-producer.sh --broker-list localhost:9092,localhost:9093,localhost:9094 --sync --topic zerg.hydra