I am using spool directory source, but am facing problem because of the buffermaxlinelength.

   My use case is, that I am storing some log files in the spool directory, however the system which is producing the log files really don't know about logging standards. There are log records of size MB. My requirement is to store these files into hadoop, so I thought of using flume. So in short whenever the spool directory source reads a record greater than the bufferMaxLineLength it throws an exception and then the process is stuck as the same message is tried over and over again. I don't want to increase the bufferMaxLineLegth to a ridiculous large value just because of very few big messages as it will require to increase the heap size (keeping the channel capacity and transaction capacity constant).
   Is there a way other than specifying befferMaxLineLength to read these unusually large messages? or any suggestions to gracefully dropping such messages while using spool directory source?