AnsweredAssumed Answered

client configuration for multiple jobtrackers

Question asked by oae on Jan 23, 2012
Latest reply on May 17, 2013 by Ted Dunning
Trying to understand the possible options for a client side configuration of a mapr cluster with multiple namenodes and jobtrackers.
From [this post][1] i know that for the namenode part i can either maintain a [mapr-clusters.conf][2] or use the MapRFileSystem which takes an array of cldb addresses.

Now the question is on the jobtrackers. Can i specify a list of them somewhere ?
I know that i could either specify 'mapred.job.tracker' with a specific host:port pair or with 'maprfs:///' and then the jobtracker host is derived from the cldb host.
But how does that work with multiple jobtrackers ? Is the current cldb host always the current jobtracker ?

Johannes
  [1]: http://answers.mapr.com/questions/935/programmatically-setting-cldb-nodes
  [2]: http://www.mapr.com/doc/display/MapR/mapr-clusters.conf

Outcomes