一些場景需要採集文字內容傳送到kafka，flume正好提供了這種解決方案

Flume agent指令碼及配置如下

Mac

安裝：

brew install flume

啟動指令碼：

nohup /usr/local/Cellar/flume/1.9.0/bin/flume-ng agent --conf /data/flume/run/flume/conf/conf.d/ --conf-file /data/flume/run/flume/conf/conf.d/mg-res_test.conf --name mg-res_test -Dflume.codeadmin.logger=INFO,console &

linux

安裝：

http://flume.apache.org/download.html

啟動指令碼：

nohup /data/flume/run/flume/bin/flume-ng agent --conf /data/flume/run/flume/conf/conf.d/ --conf-file /data/flume/run/flume/conf/conf.d/mg-res_test.conf --name mg-res_test -Dflume.codeadmin.logger=INFO,console &

配置說明

/data/flume/run/flume/conf/conf.d/mg-res_test.conf

flume主要配置資訊，關注中文描述的部分即可

# Name the components on this agent
mg-res_test.sources = mg-res_test
mg-res_test.sinks = mg-res_test
mg-res_test.channels = mg-res_test
# Describe/configure the source
mg-res_test.sources.mg-res_test.type = TAILDIR
# flume 記錄偏移量檔案
mg-res_test.sources.mg-res_test.positionFile=/data/logs/test/res-test/mg-res_test.json
mg-res_test.sources.mg-res_test.filegroups=f1
# 待採集的日誌檔案
mg-res_test.sources.mg-res_test.filegroups.f1=/data/logs/test/res-test/app.log
mg-res_test.sources.mg-res_test.fileHeader=true

# Describe the sink
mg-res_test.sinks.mg-res_test.channel = mg-res_test
mg-res_test.sinks.mg-res_test.type = org.apache.flume.sink.kafka.KafkaSink
# kafka topic配置 以實際為準
mg-res_test.sinks.mg-res_test.kafka.topic = res-test
# kafka 配置資訊 以實際為準
mg-res_test.sinks.mg-res_test.kafka.bootstrap.servers = localhost:9092
mg-res_test.sinks.mg-res_test.kafka.flumeBatchSize = 2
mg-res_test.sinks.mg-res_test.kafka.producer.acks = 1
mg-res_test.sinks.mg-res_test.kafka.producer.linger.ms = 1
mg-res_test.sinks.mg-res_test.kafka.producer.compression.type = snappy
# Use a channel which buffers events in memory
mg-res_test.channels.mg-res_test.type = memory
mg-res_test.channels.mg-res_test.capacity = 100000
mg-res_test.channels.mg-res_test.transactionCapacity = 10000
# Bind the source and sink to the channel
mg-res_test.sources.mg-res_test.channels = mg-res_test
mg-res_test.sinks.mg-res_test.channel = mg-res_test
 

/data/flume/run/flume/conf/conf.d/flume-env.sh

配置java home即可

export JAVA_OPTS="-Xms128m -Xmx128m -Dcom.sun.management.jmxremote"
JAVA_HOME=/Library/Java/JavaVirtualMachines/jdk1.8.0_192.jdk/Contents/Home

驗證

本地環境可以直接啟動kafka消費看有沒有訊息進入，測試環境請查詢kafka訊息增量或業務系統實際消費情況

kafka-console-consumer --bootstrap-server localhost:9092 --topic res-test --fro