Details
-
Sub-task
-
Status: Resolved
-
Major
-
Resolution: Fixed
-
Federation Branch
-
None
-
Reviewed
Description
The principal names in the configuration for namenode, secondary namenode, backup namenode, need not be namenode specific, i.e. they need not be configured with keys suffixed with nameserviceId.
This is because if principals were namenode specific, we would need a mechanism to figure out nameserviceId (to get the namenode specific principal) from fqdn of the files at the client.
A simpler approach is to use _HOST pattern in the principal names. Hadoop RPC supports replacing the _HOST in the principal with the hostname of the service. Using this approach we need only one principal configuration for all namenodes and similarly one for all secondary namenodes or backup namenodes.