Operating System Requirements

CDP Private Cloud Base Supported Operating Systems

Operating System support for the CDP Private Cloud Base Trial Installer

SLES 12 SP5 is not supported when using the Trial Installer (cloudera-manager-installer.bin) to install Cloudera Manager.

Operating System and IBM PowerPC support matrix

Runtime and Cloudera Manager Supported Operating Systems

Runtime provides parcels for select versions of RHEL-compatible operating systems.

Software Dependencies

• Python - CDP Private Cloud Base, with the exception of Hue, is supported on the Python version that is included in the operating system by default, as well as higher versions, but is not compatible with Python 3.0 or higher.

  For example, CDP Private Cloud Base requires Python 2.7 or higher on RHEL 7 compatible operating systems.

  Spark 2 requires Python 2.7 or higher. If the right level of Python is not picked up by default, set the PYSPARK_PYTHON and PYSPARK_DRIVER_PYTHON environment variables to point to the correct Python executable before running the pyspark command.

  Python 3 is not supported.

• Perl - Cloudera Manager requires perl.
• python-psycopg2 - Cloudera Manager 7 has a dependency on the package python-psycopg2. Hue in Runtime 7 requires a higher version of psycopg2 than is required by the Cloudera Manager dependency. For more information, see Installing the psycopg2 Python Package.
• iproute package - CDP Private Cloud Base has a dependency on the iproute package. Any host that runs the Cloudera Manager Agent requires the package. The required version varies depending on the operating system:

  Table 5. iproute package

  Operating System	iproute version
  RHEL	iproute
  UBuntu	iproute2

Filesystem Requirements

Supported Filesystems

The Hadoop Distributed File System (HDFS) is designed to run on top of an underlying filesystem in an operating system. Cloudera recommends that you use either of the following filesystems tested on the supported operating systems:

• ext3: This is the most tested underlying filesystem for HDFS.
• ext4: This scalable extension of ext3 is supported in more recent Linux releases.
• XFS: This is the default filesystem in RHEL 7.
• S3: Amazon Simple Storage Service

Kudu Filesystem Requirements - Kudu is supported on ext4 and XFS. Kudu requires a kernel version and filesystem that supports hole punching. Hole punching is the use of the fallocate(2) system call with the FALLOC_FL_PUNCH_HOLE option set.

File Access Time

Linux filesystems keep metadata that record when each file was accessed. This means that even reads result in a write to the disk. To speed up file reads, Cloudera recommends that you disable this option, called atime, using the noatime mount option in /etc/fstab:

/dev/sdb1 /data1 ext4 defaults,noatime 0

Apply the change without rebooting:

mount -o remount /data1

Filesystem Mount Options

The filesystem mount options have a sync option that allows you to write synchronously.

Using the sync filesystem mount option reduces performance for services that write data to disks, such as HDFS, YARN, Kafka and Kudu. In CDP, most writes are already replicated. Therefore, synchronous writes to disk are unnecessary, expensive, and do not measurably improve stability.

NFS and NAS options are not supported for use as DataNode Data Directory mounts, even when using Hierarchical Storage features.

Mounting /tmp as a filesystem with the noexec option is sometimes done as an enhanced security measure to prevent the execution of files stored there. However, this causes multiple problems with various parts of Cloudera Manager and CDP. Therefore, Cloudera does not support mounting /tmp with the noexec option.

Filesystem Requirements

You can control resource allocation for Cloudera Manager and CDP Runtime services (nproc, nofile, etc) from /etc/security/limits.conf, and through init scripts on traditional SysV Init systems. However, on systems using systemd the limits either needs to be set in the service’s unit file, or in /etc/systemd/system.conf, or in files present under /etc/systemd/system.conf.d/*. This is due to a known limitation with systemd as it does not use PAM login sessions (pam_limits.so) for daemon services, thereby ignoring the limits defined in /etc/security/limits.conf. Both Cloudera Manager Agent and Supervisord (responsible for starting CDP Runtime services) are daemonised during system initialisation.

nscd for Kudu

Although not a strict requirement, it's highly recommended that you usenscd to cache both DNS name resolution and static name resolution for Kudu.