Hadoop 0.20.2: Job end notification issue, JobTracker seems to suddenly stop notifying

[Wellington Chevreuil <we...@gmail.com>]

Hi Folks,

I have set a small cluster with Apache 0.20.2 distribution running in
4 machines: one as the namenode, other as the jobtracker and two
others running as both datanodes and tasktracker. I had set the
"job.end.notification.url" in mapred-site.xml, and it seems to work
fine for some time, the url set there is queried well on each job
finish. The problem is that after a while, jobtracker stops notifying
this url once jobs are finished. I could not find any error or
Exceptions on the logs, and if I restart the jobtracker, it is able to
notify again. Did anybody have already seem this issue before, or does
anyone have some tip on what could be causing this?

Thanks in advance,
Wellington.