Advertisement

Sunday, March 4, 2018

Hadoop V1 - Fair Scheduler Configuration

This is in continuation with my previous blog on Mapreduce Config
In this blog I am going to  Discuss Scheduler Configuration Fair-Scheduler -

Below is my mapred-site.xml for Fair Scheduler Configuration - 


<property>
    <name></name>
    <value></value>
</property>
<property>
    <name>mapred.jobtracker.taskScheduler</name>
    <value>org.apache.hadoop.mapred.FairScheduler</value>
</property>

<property>
    <name>mapred.fairscheduler.allocation.file</name>
    <value>/etc/hadoop/conf/fair-scheduler.xml</value>
</property>

<property>
    <name>mapred.fairscheduler.poolnameproperty</name>
    <value>user.name</value>
</property>

<property>
    <name>mapred.fairscheduler.preemption</name>
    <value>true</value>
</property>

<property>
    <name>mapred.fairscheduler.sizebasedweight</name>
    <value>true</value>
</property>

<property>
    <name>mapred.fairscheduler.assignmultiple.maps</name>
    <value>5</value>
</property>

<property>
    <name>mapred.fairscheduler.assignmultiple.reduces</name>
    <value>3</value>
</property>


Fair-Scheduler configuration File
vi /etc/hadoop/conf/fair-scheduler.xml


<allocations>
    <defaultMinSharePreemptionTimeout>600</defaultMinSharePreemptionTimeout>
    <pool name="prod-analytics">
        <minMaps>30<minMaps>
        <minReduces>10</minReduces>
    </pool>
    <pool name="dev-users">
        <weight>2</weight>
    </pool>
    <pool name="qa-users">
        <weight>1</weight>
    </pool>
    <user name="james">
        <maxRunningJobs>3</maxRunningJobs>
    </user>   

</allocations>


[As root]
# for i in $(cat /tmp/hosts) ;do scp mapred-site.xml fair-scheduler.xml ${i}:/etc/hadoop/conf/ ; done

[As root - Give Permissions[
# for i in $(cat /tmp/hosts) ;do ssh ${i} chmod -R 755 /etc/hadoop ; done;

[Stop and Restart mapred-Services]

stop-mapred.sh
start-mapred.sh


From the log file

[mapred@namenode logs]$ less hadoop-mapred-jobtracker-namenode.cluster.com.log | grep -i Fair
2018-03-03 13:43:03,169 INFO org.apache.hadoop.mapred.FairScheduler: Successfully configured FairScheduler



Verify Java Processes on all nodes

[mapred@namenode ~]$ for i in $(cat /tmp/hosts) ; do ssh ${i} 'hostname; jps | grep -vi jps; echo' ;  done;
namenode.cluster.com
4722 JobTracker


d1node.cluster.com
30419 TaskTracker

d2node.cluster.com
1600 TaskTracker

d3node.cluster.com
26777 TaskTracker

d4node.cluster.com
10144 TaskTracker

No comments:
Write comments