Adding Services

HDFS and IBM Spectrum® Scale are different file systems. If services are added only to one of the file systems, the other file system does not have the data for that service. Therefore, on switching from native HDFS to IBM Spectrum Scale or vice versa, the service cannot provide the data that you entered before switching the file system.

Ensure that you follow the User and group ids section for uid/gid consistency setup and verification.

The following are the minimum services required to be installed before you install the IBM Spectrum Scale service:
  • HDFS
  • Yarn
  • Mapreduce2
  • Zookeeper
  • SmartSense (HDP)

When adding new services to HDP, ensure that you review all configurations in the Customize Services wizard tabs. Especially, fields that set directories. This is because IBM Spectrum Scale is a shared file system. Ensure that any services do not use a shared file system mount point when it requires a local directory. See Create HDP cluster>Installing, Configuring, and Deploying a Cluster >Customize Services>Directories section.

If the service is already added with the remote mount point set as one of the directories, and you want to modify or remove the remote mount point directory from the service, ensure that you restart the service to pick up the new configuration changes.

If you are planning to install High Availability (HA), ensure that you setup the HA in native HDFS mode. Do not install HA when IBM Spectrum Scale Service is integrated.

If IBM Spectrum Scale is integrated, see HDFS NameNode High Availability [HA].