Баг на баге и багом погоняет :-)

Из исходников Flume

1. // (1) At the time of this writing, Hadoop has a bug which causes the
2. // non-native gzip implementation to create invalid gzip files when
3. // finish() and resetState() are called. See HADOOP-8522.
4. //
5. // (2) Even if HADOOP-8522 is fixed, the JDK GZipInputStream is unable
6. // to read multi-member (concatenated) gzip files. See this Sun bug:
7. // http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=4691425