Mappers fail to initialize and are killed after 600 seconds

Hi all, I'm running a simple mapreduce job that connects to an hbase table, reads each row, counts some co-occurrence frequencies, and writes everything out to hdfs at the end. Everything seems to be going smoothly until the last 5, out of 108, tasks run. The last 5 tasks seem to be stuck

Re: Mappers fail to initialize and are killed after 600 seconds

Are there no userlogs from the failed tasks? TaskTracker logs won't carry user-code (task) logs. Could you paste those syslog lines (from the task) to pastebin/etc. since the lists may not be accepting attachments? On Tue, Aug 2, 2011 at 12:51 AM, Stevens, Keith D. steven...@llnl.gov wrote: Hi

Re: Mappers fail to initialize and are killed after 600 seconds

In short, there are no userlogs. stderr and stdout are both empty. I copied the output from syslog to the following pastebin: http://pastebin.com/0XXE9Jze. The first 22 lines look to be exactly the same as the syslogs for other, non-dying, tasks. The main departure is on line 23 where the