Running Hadoop MapReduce word count for the first time fails? - hadoop

Question

When running the Hadoop word count example the first time it fails. Here's what I'm doing:
Format namenode: $HADOOP_HOME/bin/hdfs namenode -format
Start HDFS/YARN:
$HADOOP_HOME/sbin/start-dfs.sh
$HADOOP_HOME/sbin/start-yarn.sh
$HADOOP_HOME/sbin/yarn-daemon.sh start nodemanager
Run wordcount: hadoop jar $HADOOP_HOME/share/hadoop/mapreduce/hadoop-mapreduce-examples-2.7.2.jar wordcount input output
(let's say input folder is already in HDFS I'm not gonna put every single command here)
Output:
16/07/17 01:04:34 INFO client.RMProxy: Connecting to ResourceManager at hadoop-master/172.20.0.2:8032
16/07/17 01:04:35 INFO input.FileInputFormat: Total input paths to process : 2
16/07/17 01:04:35 INFO mapreduce.JobSubmitter: number of splits:2
16/07/17 01:04:36 INFO mapreduce.JobSubmitter: Submitting tokens for job: job_1468688654488_0001
16/07/17 01:04:36 INFO impl.YarnClientImpl: Submitted application application_1468688654488_0001
16/07/17 01:04:36 INFO mapreduce.Job: The url to track the job: http://hadoop-master:8088/proxy/application_1468688654488_0001/
16/07/17 01:04:36 INFO mapreduce.Job: Running job: job_1468688654488_0001
16/07/17 01:04:46 INFO mapreduce.Job: Job job_1468688654488_0001 running in uber mode : false
16/07/17 01:04:46 INFO mapreduce.Job: map 0% reduce 0%
Terminated
And then HDFS crashes so I can't access http://localhost:50070/
Then I restart eveyrthing (repeat step 2), rerun the example and everything's fine.
How can I fix it for the first run? My HDFS obviously has no data the first time around, maybe that's the problem?
UPDATE:
Running an even simpler example fails as well:
hadoop#8f98bf86ceba:~$ hadoop jar $HADOOP_HOME/share/hadoop/mapreduce/hadoop-mapreduce-examples*.jar pi 3 3
Number of Maps = 3
Samples per Map = 3
Wrote input for Map #0
Wrote input for Map #1
Wrote input for Map #2
Starting Job
16/07/17 03:21:28 INFO client.RMProxy: Connecting to ResourceManager at hadoop-master/172.20.0.3:8032
16/07/17 03:21:29 INFO input.FileInputFormat: Total input paths to process : 3
16/07/17 03:21:29 INFO mapreduce.JobSubmitter: number of splits:3
16/07/17 03:21:29 INFO mapreduce.JobSubmitter: Submitting tokens for job: job_1468696855031_0001
16/07/17 03:21:31 INFO impl.YarnClientImpl: Submitted application application_1468696855031_0001
16/07/17 03:21:31 INFO mapreduce.Job: The url to track the job: http://hadoop-master:8088/proxy/application_1468696855031_0001/
16/07/17 03:21:31 INFO mapreduce.Job: Running job: job_1468696855031_0001
16/07/17 03:21:43 INFO mapreduce.Job: Job job_1468696855031_0001 running in uber mode : false
16/07/17 03:21:43 INFO mapreduce.Job: map 0% reduce 0%
Same problem, HDFS terminates

Answer 1

Your post looks incomplete to deduce what is wrong here. My guess is that hadoop-mapreduce-examples-2.7.2-sources.jar is not what you want. More likely you need hadoop-mapreduce-examples-2.7.2.jar containing .class files and not the sources.

Answer 2

HDFS has to be restarted the first time before MapReduce jobs can be successfully ran. This is because HDFS creates some data on the first run but stopping it can clean up its state so MapReduce jobs can be ran through YARN afterwards.
So my solution was:
Start Hadoop: $HADOOP_HOME/sbin/start-dfs.sh
Stop Hadoop: $HADOOP_HOME/sbin/stop-dfs.sh
Start Hadoop again: $HADOOP_HOME/sbin/start-dfs.sh