How to separate Hadoop MapReduce from HDFS? -
i'm curious if separate hdfs filesystem mapreduce framework. know main point of hadoop run maps , reduces on machines data in question, wondering if change *.xml files change configuration of machine jobtracker, namenode , datanodes running on.
currently, configuration 2 vms setup: 1 (the master) namenode, datanode, jobtracker, tasktracker (and secondarynamenode), other (the slave) datanode, tasktraker. essentially, want change have master namenode datanode(s), jobtracker, , have slave tasktracker perform computations (and later on, have more slaves tasktrackers on them; 1 on each). bottleneck data transfer between 2 vms computations of maps , reduces, since data @ stage small i'm not concerned it. know if configuration possible, , how it. tips?
thanks!
you don't specify kind of options in configuration files. have take care of kind of deamons start on each machine(you call them vms think mean machines).
i suppose start using start-all.sh script can find in bin directory under hadoop installation dir.
if take @ script see call number of sub-scripts corresponding starting datanodes, tasktrackers , namenode, jobtracker.
in order achive you've said, this:
modify masters , slaves files this: master file should contain name of machine1 slaves should contain name of machine2
run start-mapred.sh
modify masters , slaves files this: master file should contain machine1 slaves file should contain machine1
run start-dfs.sh
i have tell i've never tried such configuration i'm not sure going work can give try. anyway solution in direction!
Comments
Post a Comment