On Mi 15 feb 2012 00:10:52 +0200, Steve Lewis wrote: > All of my current jobs and the wordcount I used when learning are all > failing with the same error > > : Error reading task outputhttp:// > glados9.systemsbiology.net:50060/tasklog?plaintext=true&taskid=attempt_201202141331_0008_m_000001_2&filter=stdout > > glados9 is a slave node and other slave nodes are listed as well > > I have tried deleting the userlogs directory of all slave nodes without > success - > I see no useful logs and am at a real loss of what to do I suspect some > directory somewhere has too many entries but other than userlogs I am not > sure where else to look
I've also seen that output, and it was related to the job failing to run (I think it was due to a different version of slf4j on the classpath; make sure you use the one hadoop is compiled with or a compatible one). Check the task tracker and job tracker logs on the data node where the job ran to see why it failed.
