Virtual machine configuration

• RAM: 4 GB
• Disk: 100 GB (only uses what it needs)

While following the instructions from SkillSoft on how to perform a Hadoop installation on a virtual CentOS virtual machine, an issue occurred. After configuring HDFS and successfully formatting my HDFS file system, it was time to start the HDFS daemon.

Per the instructions, the command used was:

  hadoop-daemon.sh start namenode
  

A large number of error messages were returned, so I started at the top with the first one:

  mkdir: cannot create directory '/var/run/hadoop-hdfs': Permission denied
  

I remembered an anomaly occurring during the HDFS setup. When I issue the command:

  ls -l /var/run
  

The output that I receive says:

  lrwxrwxrwx. 1 root root 6 Dec 18 16:10 /var/run -> ../run
  

This means that the /var/run directory is actually a symbolic link to the /run directory. When I look at the /run directory, there is no file or directory called hadoop-hdfs.

I need a directory called /run/hadoop-hdfs with ownership going to hduser:hadoop. So I did the following:

  sudo mkdir /run/hadoop-hdfs
  sudo chown -R hduser:hadoop /var/run/hadoop-hdfs
  sudo chmod 775 /var/run/hadoop-hdfs