第二种sparkStreaming 整合Flume

flume采用 netcat-memory-customer sink架构

本地测试

1：本地启动sprakStreaming服务，（192.168.145.128 10000）

2. 服务器中启动flume agent

3. telnet往端口中输入数据，观察本地idea控制台输出数据

服务器测试

mvn打包：mvn clean package -DskipTests

上传至服务器

先启动flume

flume-ng agent \
  --name netcat-memory-avro \
  --conf $FLUME_HOME/conf \
  --conf-file $FLUME_HOME/conf/netcat-memory-avro.conf \
  -Dflume.root.logger=INFO,console 

后启动spark

spark-submit \
--class com.tuzhihai.flumespark.SparkPullFlume \
--master local[2] \
--packages org.apache.spark:spark-streaming-flume_2.11:2.2.0 \
/root/soft_down/lib/sparklearn-1.0.jar \
192.168.145.128 10000

在端口输入数据

telnet 192.168.145.128 9990

观察flume控制台

pull方式为什么要先启动flume后启动spark

首先说明，pull方式比push方式更加可靠，在实际工作中应用极多。

pull方式是flume采集到数据后存储在一个Agent中，然后spark想要数据的时候，直接从这个Agent中pull就ok了。这个方式明显更加友好，更加符合工作要求，不是你主动传给我，而是我想要的时候我自己从你那拿。

spark-pull-flume.conf

 flume-ng agent \
  --name netcat-memory-spark \
  --conf $FLUME_HOME/conf \
  --conf-file $FLUME_HOME/conf/spark-pull-flume.conf \
  -Dflume.root.logger=INFO,console 

# example netcat-memory-spark
netcat-memory-spark.sources = netcat-source
netcat-memory-spark.sinks = spark-sink
netcat-memory-spark.channels = memory-channel

# Describe/configure the source
netcat-memory-spark.sources.netcat-source.type = netcat
netcat-memory-spark.sources.netcat-source.bind = 192.168.145.128
netcat-memory-spark.sources.netcat-source.port = 9999

# Describe/ the sink
netcat-memory-spark.sinks.spark-sink.type = org.apache.spark.streaming.flume.sink.SparkSink
netcat-memory-spark.sinks.spark-sink.hostname = 192.168.145.128
netcat-memory-spark.sinks.spark-sink.port = 10000

# Use a channel which buffers events in memory
netcat-memory-spark.channels.memory-channel.type = memory

# Bind the source and sink to the channel
netcat-memory-spark.sources.netcat-source.channels = memory-channel
netcat-memory-spark.sinks.spark-sink.channel = memory-channel