Yes it would be very useful to have more information. Regarding your configuration, the most common setup is to put your Namenode, HMaster, and JobTracker on a single node (this node becomes the only single point of failure).
You would then have three matching servers running datanode, regionserver, tasktracker. One very important factor is the configuration of your tasktrackers. What do you have set as maximum maps/reduces per tasktracker? (in hadoop conf, mapred.tasktracker.map.tasks.maximum and mapred.tasktracker.reduce.tasks.maximum). With the hbase master co-hosted with a regionserver, datanode, tasktracker, as well as a running MR job, it would not be surprising to see something go wrong. With 4 cores, if you had that server setup with all four on one server, you'd only want a single map or single reduce running at any one time to prevent starvation. As JD said, do you have any logs? Please turn your hbase logging to DEBUG (in log4j.properties) and send us the logs from the crash. You also might change your configuration as above and rerun again with DEBUG. JG > -----Original Message----- > From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Jean- > Daniel Cryans > Sent: Tuesday, November 25, 2008 7:38 AM > To: [email protected] > Subject: Re: configuration fo map reduce > > What kind of crash? Do you have a stack trace from the logs? What kind > of MR > job was that (source, size, etc)? What's your current configuration > (*-site.xml)? > > Without any of these it would be hard to assert anything. > > J-D > > 2008/11/25 ROL <[EMAIL PROTECTED]> > > > Hi. > > > > I have hbase master crashed after BulkImport mr job; > > > > is it possible that's because my configuration is bad? Can you > > suggest configuration for map reduce jobs with that: > > > > -hadoop 0.18.2 > > -hbase 0.18.1 > > > > 4 nodes: > > > > 4CPU, 4Gb RAM > > > > > > my current configuration: > > > > 1 node : namenode, datanode, tasktracker, hbase region > > 2 node : jobtracker, datanode, tasktracker, hbase region > > 3 node : hbase master, datanode, tasktracker, hbase rehion > > 4 node : datanode, tasktracker, hbase region > > > > > > > > -- > > С уважением, > > ROL mailto:[EMAIL PROTECTED] > > > >
