flume SpoolDir hdfs sink

Below is my Flume config file to push files dropped in folder to HDFS
The files are usually about 2MB in size.
The default property deserializer.maxLineLength is set to 2048. Which means after 2048 bytes of data, 
flume truncates the data and treats it as a new event. Thus the resulting file in HDFS had a lot of newlines.
I changed it to 4096000, which is about 4MB



#Flume config file
tier1.sources = xml-source1
tier1.channels = channel1
tier1.sinks = hdfs-sink1

tier1.channels.channel1.type = memory
tier1.channels.channel1.checkpointDir = /mnt/data/tenny/checkpoint_xml
tier1.channels.channel1.dataDirs = /mnt/data/tenny/kafka_source_xml
tier1.channels.channel1.capacity = 10000000
tier1.channels.channel1.transactionCapacity = 100000


tier1.sources.xml-source1.type = spooldir
tier1.sources.xml-source1.channels = channel1
tier1.sources.xml-source1.spoolDir = /mnt/data/tenny/SpoolDir
tier1.sources.xml-source1.fileHeader = true
tier1.sources.xml-source1.fileSuffix = .DONE
tier1.sources.xml-source1.basenameHeader = true
tier1.sources.xml-source1.ignorePattern = .*.tmp$
tier1.sources.xml-source1.deserializer.maxLineLength = 4096000

tier1.sinks.hdfs-sink1.type = hdfs
tier1.sinks.hdfs-sink1.hdfs.path = /user/tsusanto/salesfiles/%Y%m%d/%H00
tier1.sinks.hdfs-sink1.brokerList = broker1_host:9092,broker2_host:9092,broker3_host:9092
tier1.sinks.hdfs-sink1.channel = channel1
tier1.sinks.hdfs-sink1.hdfs.useLocalTimeStamp = true
tier1.sinks.hdfs-sink1.hdfs.rollCount = 0
tier1.sinks.hdfs-sink1.hdfs.rollInterval = 5
tier1.sinks.hdfs-sink1.hdfs.rollSize = 0
tier1.sinks.hdfs-sink1.hdfs.fileType = DataStream
tier1.sinks.hdfs-sink1.hdfs.writeFormat = Text
tier1.sinks.hdfs-sink1.hdfs.filePrefix = %{basename}