workflow jobs stuck in prep state on single manager and 3 data node cluster.

I have a HDP cluster with single manager node and 3 data nodes. workflow job's status shows as running but the task status shows a prep and it is stuck in the same state forever.

The same job works fine in an HDP cluster with 3 manager and 4 data node setup. 

I have also verified that the jobs.properties is pointing to proper job tracker and nameNode.

There is nothing unusual or any relevant warning seen in the oozie.log or in oozie-error.log.

There is no resource crunch in Yarn too, as there are no other jobs using yarn's resources.

oozie status too says 'Normal'

I have recreated the setup multiple times and the issue is consistent.