I ended up having a cname record created for my master server (cluster-00) and when putting that in hadoop-site.xml everything worked fine. Before it was simply the machine name (nutch). I'm not quite sure whether it was our DNS servers confusing things or something on the slave machines, or something within nutch, but it is fixed now. Thanks for the help.
Jeff -----Original Message----- From: Dennis Kubes [mailto:[EMAIL PROTECTED] Sent: Wednesday, December 19, 2007 11:12 PM To: [email protected] Subject: Re: cluster connectivity Sounds like the slaves namenode and/or jobtracker configuration is pointing to localhost in the hadoop-site.xml file. Dennis Kubes Bolle, Jeffrey F. wrote: > I have a rather odd problem. > > We had some DNS issues and the DNS servers were reloaded over the > weekend. > > Now, my nutch cluster can't see itself. I can easily ssh between > machines, name resolution seems to be working just fine, they can all > ping each other, etc. The problem is that when I run bin/start-all.sh > and check the web cluster summary only one node is connected (the slave > node started on the master machine). If I ssh into one of the slave > nodes and check the logs it is trying to connect to the master node, > but to no avail. Does anyone have any recommendations on where things > are messed up. Oh, I should add that everything properly rsyncs and > starts when start-all.sh is run, the only thing that doesn't happen is > the slaves connecting back to the master. > > Thanks for any help. > > Jeff > >
