[Hadoop][AWS] Mapreduce WordCount 실행 시, os::commit_memory(0x00000000f660c000, 104861696, 0) failed; error='Cannot allocate memory' (errno=12) 에러 발생 시 해결 방법

 

Hadoop 스터디를 위해 설치한 AWS 서버가 free teer 버전이라 메모리가 1G 였다.

따라서 Hadoop 설치 후, Mapreduce의 Wordcount를 실행하려하니 아래와 같은 에러가 발생했다.

 

1. 에러

 

 

OpenJDK 64-Bit Server VM warning: INFO: os::commit_memory(0x00000000f660c000, 104861696, 0) failed; error='Cannot allocate memory' (errno=12) # There is insufficient memory for the Java Runtime Environment to continue. # Native memory allocation (mmap) failed to map 104861696 bytes for committing reserved memory. # An error report file with more information is saved as: # /home/hadoop/hadoop/hs_err_pid3970.log[hadoop@ip-172-31-33-183 hadoop]$ hadoop jar share/hadoop/mapreduce/hadoop-mapreduce-examples-3.1.2.jar wordcount input output 2020-03-27 16:07:02,698 WARN util.NativeCodeLoader: Unable to load native-hadoop library for your platform... using builtin-java classes where applicable 2020-03-27 16:07:03,994 INFO impl.MetricsConfig: loaded properties from hadoop-metrics2.properties 2020-03-27 16:07:04,211 INFO impl.MetricsSystemImpl: Scheduled Metric snapshot period at 10 second(s). 2020-03-27 16:07:04,211 INFO impl.MetricsSystemImpl: JobTracker metrics system started 2020-03-27 16:07:04,689 INFO input.FileInputFormat: Total input files to process : 1 2020-03-27 16:07:04,752 INFO mapreduce.JobSubmitter: number of splits:1 2020-03-27 16:07:05,043 INFO mapreduce.JobSubmitter: Submitting tokens for job: job_local266812600_0001 2020-03-27 16:07:05,059 INFO mapreduce.JobSubmitter: Executing with tokens: [] 2020-03-27 16:07:05,341 INFO mapreduce.Job: The url to track the job: http://localhost:8080/ 2020-03-27 16:07:05,342 INFO mapreduce.Job: Running job: job_local266812600_0001 2020-03-27 16:07:05,358 INFO mapred.LocalJobRunner: OutputCommitter set in config null 2020-03-27 16:07:05,378 INFO output.FileOutputCommitter: File Output Committer Algorithm version is 2 2020-03-27 16:07:05,378 INFO output.FileOutputCommitter: FileOutputCommitter skip cleanup _temporary folders under output directory:false, ignore cleanup failures: false 2020-03-27 16:07:05,379 INFO mapred.LocalJobRunner: OutputCommitter is org.apache.hadoop.mapreduce.lib.output.FileOutputCommitter 2020-03-27 16:07:05,471 INFO mapred.LocalJobRunner: Waiting for map tasks 2020-03-27 16:07:05,472 INFO mapred.LocalJobRunner: Starting task: attempt_local266812600_0001_m_000000_0 2020-03-27 16:07:05,496 INFO output.FileOutputCommitter: File Output Committer Algorithm version is 2 2020-03-27 16:07:05,496 INFO output.FileOutputCommitter: FileOutputCommitter skip cleanup _temporary folders under output directory:false, ignore cleanup failures: false 2020-03-27 16:07:05,527 INFO mapred.Task:  Using ResourceCalculatorProcessTree : [ ] 2020-03-27 16:07:05,530 INFO mapred.MapTask: Processing split: hdfs://localhost:9000/user/hadoop/input/LICENSE.txt:0+147145 OpenJDK 64-Bit Server VM warning: INFO: os::commit_memory(0x00000000f660c000, 104861696, 0) failed; error='Cannot allocate memory' (errno=12) # # There is insufficient memory for the Java Runtime Environment to continue. # Native memory allocation (mmap) failed to map 104861696 bytes for committing reserved memory. # An error report file with more information is saved as: # /home/hadoop/hadoop/hs_err_pid3970.log

 

 

2. 해결 방법

Swap 메모리를 추가해주기로 했다.

 

https://aws.amazon.com/ko/premiumsupport/knowledge-center/ec2-memory-swap-file/ sudo dd if=/dev/zero of=/swapfile bs=128M count=32 sudo chmod 600 /swapfile sudo mkswap /swapfile sudo swapon /swapfile sudo swapon -s sudo vi /etc/fstab [추가] /swapfile swap swap defaults 0 0

 

free -h 로 메모리가 증가되었는지 확인해보자.

 

 

[추가 전]

 

[추가 후]

 

이 후 wordcount를 실행해보면 정상 종료 되는 것을 확인 할 수 있다.

 

 

* 'java.net.ConnectException: Call From ip-172-31-33-183.ap-northeast-2.compute.internal/172.31 .33.183 to localhost:9000 failed on connection exception' 에러가 발생할 경우

 

[hadoop@ip-172-31-33-183 hadoop]$ hadoop jar share/hadoop/mapreduce/hadoop-mapreduce-examples-3.1.2.jar wordcount input output 2020-03-27 16:39:39,705 WARN util.NativeCodeLoader: Unable to load native-hadoop library for your platform... using builtin-java classes where applicable 2020-03-27 16:39:41,118 INFO impl.MetricsConfig: loaded properties from hadoop-metrics2.properties 2020-03-27 16:39:41,281 INFO impl.MetricsSystemImpl: Scheduled Metric snapshot period at 10 second(s). 2020-03-27 16:39:41,281 INFO impl.MetricsSystemImpl: JobTracker metrics system started java.net.ConnectException: Call From ip-172-31-33-183.ap-northeast-2.compute.internal/172.31 .33.183 to localhost:9000 failed on connection exception: java.net.ConnectException: Connection refused; For more details see:  http://wiki.apache.org/hadoop/ConnectionRefused         at sun.reflect.NativeConstructorAccessorImpl.newInstance0(Native Method)         at sun.reflect.NativeConstructorAccessorImpl.newInstance(NativeConstructorAccessorImpl.java:62)         at sun.reflect.DelegatingConstructorAccessorImpl.newInstance(DelegatingConstructorAccessorImpl.java:45)         at java.lang.reflect.Constructor.newInstance(Constructor.java:423)         at org.apache.hadoop.net.NetUtils.wrapWithMessage(NetUtils.java:831)         at org.apache.hadoop.net.NetUtils.wrapException(NetUtils.java:755)         at org.apache.hadoop.ipc.Client.getRpcResponse(Client.java:1515)

 

 

원인은 아까 swap 메모리 설치 시, exit를 통해 job을 비정상 종료했기 때문이다.

정상적으로 모든 job이 떠있는지 jps를 통해 확인해보자.

 

 

start-all.sh

 

startp-all.sh를 통해 namenode가 start되는 것을 확인한 후, 실행하면 정상 실행 된다.