在使用过cdh6.3.1版本的flume时，配置了从kafka消费写入hdfs，使用的是FileChannel。
使用FileChannel还是考虑到数据可靠性。
但是今天在配置的时候，一直不生效。
由于我在配置FileChannel时，在指定的Agent的服务器上去设置目录。
这里必须将该目录的owner用户设置为hdfs，否则就会导致不生效。但是在后台的日志信息中看不到任何消息的。
chown -R hdfs:hdfs /home/hadoop/bigdata/flume_job
我的agent配置如下：
tier1.sources=bsc_traces uni_v3
tier1.sinks=bsc_traces uni_v3
tier1.channels=bsc_traces uni_v3
tier1.sources.bsc_traces.type=org.apache.flume.source.kafka.KafkaSource
tier1.sources.bsc_traces.batchSize=100
tier1.sources.bsc_traces.batchDurationMillis=3000
tier1.sources.bsc_traces.kafka.bootstrap.servers=192.168.1.17:9092
tier1.sources.bsc_traces.kafka.topics=testTopic2
tier1.sources.bsc_traces.kafka.consumer.group.id=bigdata_ods_bsc_traces
tier1.sources.bsc_traces.kafka.consumer.auto.offset.reset=earliest
tier1.sources.bsc_traces.kafka.consumer.auto.commit.enable=false
tier1.sources.bsc_traces.kafka.consumer.timeout.ms=15000
tier1.sources.bsc_traces.kafka.consumer.fetch.max.wait.ms=5000
tier1.sources.bsc_traces.kafka.consumer.max.poll.records=100
tier1.sources.bsc_traces.kafka.consumer.max.poll.interval.ms=3000000
tier1.sources.bsc_traces.interceptors=i1
tier1.sources.bsc_traces.interceptors.i1.type=com.xxx.flume.interceptor.AutoPartitionInterceptor$Builder
tier1.sources.bsc_traces.interceptors.i1.timestampField=block_timestamp:GMT+0
tier1.sources.bsc_traces.channels=bsc_traces
tier1.sinks.bsc_traces.type=hdfs
tier1.sinks.bsc_traces.hdfs.path=hdfs://ns1/user/hive/warehouse/ods.db/ods_bsc_traces/pk_year=%{pk_year}/pk_month=%{pk_month}/pk_day=%{pk_day}
tier1.sinks.bsc_traces.hdfs.filePrefix=ods_bsc_traces
tier1.sinks.bsc_traces.hdfs.fileSufix=.log
tier1.sinks.bsc_traces.hdfs.useLocalTimeStamp=true
tier1.sinks.bsc_traces.hdfs.batchSize=500
tier1.sinks.bsc_traces.hdfs.fileType=DataStream
tier1.sinks.bsc_traces.hdfs.writeFormat=Text
tier1.sinks.bsc_traces.hdfs.rollSize=2147483648
tier1.sinks.bsc_traces.hdfs.rollInterval=0
tier1.sinks.bsc_traces.hdfs.rollCount=0
tier1.sinks.bsc_traces.hdfs.idleTimeout=120
tier1.sinks.bsc_traces.hdfs.minBlockReplicas=1
tier1.sinks.bsc_traces.channel=bsc_traces
tier1.channels.bsc_traces.type=file
tier1.channels.bsc_traces.checkpointDir=/home/hadoop/bigdata/flume_job/chkDir/ods_bsc_traces
tier1.channels.bsc_traces.dataDirs=/home/hadoop/bigdata/flume_job/dataDir/ods_bsc_traces
tier1.sources.uni_v3.type=org.apache.flume.source.kafka.KafkaSource
tier1.sources.uni_v3.batchSize=100
tier1.sources.uni_v3.batchDurationMillis=3000
tier1.sources.uni_v3.kafka.bootstrap.servers=192.168.1.17:9092
tier1.sources.uni_v3.kafka.topics=TestTopic
tier1.sources.uni_v3.kafka.consumer.group.id=bigdata2_uni_v3
tier1.sources.uni_v3.kafka.consumer.auto.offset.reset=earliest
tier1.sources.uni_v3.kafka.consumer.auto.commit.enable=false
tier1.sources.uni_v3.kafka.consumer.timeout.ms=15000
tier1.sources.uni_v3.kafka.consumer.fetch.max.wait.ms=5000
tier1.sources.uni_v3.kafka.consumer.max.poll.records=100
tier1.sources.uni_v3.kafka.consumer.max.poll.interval.ms=3000000
tier1.sources.uni_v3.interceptors=i1
tier1.sources.uni_v3.interceptors.i1.type=com.xxx.flume.interceptor.BlockExtractorInterceptor$Builder
tier1.sources.uni_v3.channels=uni_v3
tier1.sinks.uni_v3.type=hdfs
tier1.sinks.uni_v3.hdfs.path=hdfs://ns1/user/hive/warehouse/ods.db/ods_%{project}_%{name}/pk_year=%{pk_year}/pk_month=%{pk_month}/pk_day=%{pk_day}
tier1.sinks.uni_v3.hdfs.filePrefix=%{project}_%{name}
tier1.sinks.uni_v3.hdfs.fileSufix=.log
tier1.sinks.uni_v3.hdfs.useLocalTimeStamp=true
tier1.sinks.uni_v3.hdfs.batchSize=500
tier1.sinks.uni_v3.hdfs.fileType=DataStream
tier1.sinks.uni_v3.hdfs.writeFormat=Text
tier1.sinks.uni_v3.hdfs.rollSize=2147483648
tier1.sinks.uni_v3.hdfs.rollInterval=3600
tier1.sinks.uni_v3.hdfs.rollCount=0
tier1.sinks.uni_v3.hdfs.idleTimeout=120
tier1.sinks.uni_v3.hdfs.minBlockReplicas=1
tier1.sinks.uni_v3.channel=uni_v3
tier1.channels.uni_v3.type=file
tier1.channels.uni_v3.checkpointDir=/home/hadoop/bigdata/flume_job/chkDir/uni_v3
tier1.channels.uni_v3.dataDirs=/home/hadoop/bigdata/flume_job/dataDir/uni_v3