Jobs fail in Yarn with out of Java heap memory error

[#1]

We are running Yarn on CDH 5.1 with 14 nodes using 6 GB of memory. I understand this is not a log of memory, but it is all we could put together. Most jobs complete without error, but a few of the larger MapReduce jobs fail with an out of Java heap memory error. The jobs fail on a Reduce task that either sorts or groups data. We recently upgraded to CDH 5.1 from CDH 4.7 and ALL of these jobs succeeded on MapReduce v1. Looking in the logs I see that the Application has retired a few times before failing. Can you see anything wrong with the way the resources are configured?

Java Heap Size of NodeManager in Bytes	1 GB
yarn.nodemanager.resource.memory-mb	6 GB
yarn.scheduler.minimum-allocation-mb	1 GB
yarn.scheduler.maximum-allocation-mb	6 GB
yarn.app.mapreduce.am.resource.mb	1.5 GB
yarn.nodemanager.container-manager.thread-count	20
yarn.resourcemanager.resource-tracker.client.thread-count	20
mapreduce.map.memory.mb	1.5 GB
mapreduce.reduce.memory.mb	3 GB
mapreduce.map.java.opts	"-Djava.net.preferIPv4Stack=true -Xmx 1228m";
mapreduce.reduce.java.opts	"-Djava.net.preferIPv4Stack=true -Xmx2457m";
mapreduce.task.io.sort.factor	5
mapreduce.task.io.sort.mb	512 MB
mapreduce.job.reduces	2
mapreduce.reduce.shuffle.parallelcopies	4

One thing that might help, Yarn runs 4 containers per node, can this be reduced?

[#2]

I'm confused. Your initial post says that the reduce heap is 2457MB. Now it seems that's just 787.69MB. Which one is right? What does /etc/hadoop/conf/mapred-site.xml say?

[#3]

Here is what I have. Is the map/reduce java opts being overwritten, there are two entries? 

<!--Autogenerated by Cloudera Manager-->
<configuration>
<property>
<name>mapreduce.job.split.metainfo.maxsize</name>
<value>10000000</value>
</property>
<property>
<name>mapreduce.job.counters.max</name>
<value>120</value>
</property>
<property>
<name>mapreduce.output.fileoutputformat.compress</name>
<value>false</value>
</property>
<property>
<name>mapreduce.output.fileoutputformat.compress.type</name>
<value>BLOCK</value>
</property>
<property>
<name>mapreduce.output.fileoutputformat.compress.codec</name>
<value>org.apache.hadoop.io.compress.DefaultCodec</value>
</property>
<property>
<name>mapreduce.map.output.compress.codec</name>
<value>org.apache.hadoop.io.compress.SnappyCodec</value>
</property>
<property>
<name>mapreduce.map.output.compress</name>
<value>true</value>
</property>
<property>
<name>zlib.compress.level</name>
<value>DEFAULT_COMPRESSION</value>
</property>
<property>
<name>mapreduce.task.io.sort.factor</name>
<value>5</value>
</property>
<property>
<name>mapreduce.map.sort.spill.percent</name>
<value>0.8</value>
</property>
<property>
<name>mapreduce.reduce.shuffle.parallelcopies</name>
<value>4</value>
</property>
<property>
<name>mapreduce.task.timeout</name>
<value>600000</value>
</property>
<property>
<name>mapreduce.client.submit.file.replication</name>
<value>4</value>
</property>
<property>
<name>mapreduce.job.reduces</name>
<value>2</value>
</property>
<property>
<name>mapreduce.task.io.sort.mb</name>
<value>512</value>
</property>
<property>
<name>mapreduce.map.speculative</name>
<value>false</value>
</property>
<property>
<name>mapreduce.reduce.speculative</name>
<value>false</value>
</property>
<property>
<name>mapreduce.job.reduce.slowstart.completedmaps</name>
<value>0.8</value>
</property>
<property>
<name>mapreduce.jobhistory.address</name>
<value>blvdevhdp05.ds-iq.corp:10020</value>
</property>
<property>
<name>mapreduce.jobhistory.webapp.address</name>
<value>blvdevhdp05.ds-iq.corp:19888</value>
</property>
<property>
<name>mapreduce.framework.name</name>
<value>yarn</value>
</property>
<property>
<name>yarn.app.mapreduce.am.staging-dir</name>
<value>/user</value>
</property>
<property>
<name>yarn.app.mapreduce.am.resource.mb</name>
<value>1536</value>
</property>
<property>
<name>yarn.app.mapreduce.am.resource.cpu-vcores</name>
<value>1</value>
</property>
<property>
<name>mapreduce.job.ubertask.enabled</name>
<value>false</value>
</property>
<property>
<name>yarn.app.mapreduce.am.command-opts</name>
<value>-Djava.net.preferIPv4Stack=true -Xmx825955249</value>
</property>
<property>
<name>mapreduce.map.java.opts</name>
<value>-Djava.net.preferIPv4Stack=true -Xmx768m -Xmx825955249</value>
</property>
<property>
<name>mapreduce.reduce.java.opts</name>
<value>-Djava.net.preferIPv4Stack=true -Xmx1280m -Xmx825955249</value>
</property>
<property>
<name>mapreduce.map.memory.mb</name>
<value>1280</value>
</property>
<property>
<name>mapreduce.map.cpu.vcores</name>
<value>1</value>
</property>
<property>
<name>mapreduce.reduce.memory.mb</name>
<value>1792</value>
</property>
<property>
<name>mapreduce.reduce.cpu.vcores</name>
<value>1</value>
</property>
<property>
<name>mapreduce.application.classpath</name>
<value>$HADOOP_MAPRED_HOME/*,$HADOOP_MAPRED_HOME/lib/*,$MR2_CLASSPATH</value>
</property>
<property>
<name>mapreduce.admin.user.env</name>
<value>LD_LIBRARY_PATH=$HADOOP_COMMON_HOME/lib/native:$JAVA_LIBRARY_PATH</value>
</property>
<property>
<name>mapreduce.shuffle.max.connections</name>
<value>80</value>
</property>
</configuration>

[#4]

I have an easy question, if I increase mapreduce.reduce.shuffle.parallelcopies from 4 to 10, will that increase or decrease memory used by the node?

It seems to me that if this is increased that data is written quickly to files and out of memory. But I might be wrong...

[#5 · Accepted solution]

<name>mapreduce.reduce.java.opts</name>
<value>-Djava.net.preferIPv4Stack=true -Xmx1280m -Xmx825955249</value>

Limits the heap to ~825MB. Most JVMs resolve duplicate args by picking the last one. So this is nowhere close to the 3GB that you intended. You should find out where you set this in CM and change it.

Do that before you play with parallelcopies. But to answer your questions, yes, it'll increase CPU, memory & network usage. And it could lead to more disk spills and slow down your job.

[#6]

You pointed out the problem and I removed the -Xmx825955249 from where I had entered it in Cloudera Manager. I was using the wrong field to update the value. Thank you so much for sticking with me and helping me resolve this issue! The jobs now succeed!

Kevin Verhoeven