what is the ideal number of nodes(how many masters and how many slaves) should be probation in hadoop production environment?

[#1]

@Barada Nayak To acurately answer your question we need more info.

Which services do you plan to use (hive, oozie, hbase, etc)

Is HA required for all services? This is a business decision on HA requirements

what are the data volumes the cluster with host?

what are the mutation figures?

Generally you will find small clusters with 3 masters, medium with 4 and large with at least 5. It all depend on your use case. I have seen in the field where name node gets pounded and I have recommended to isolate the name node (not share with other master services).

[#2]

Thanks @Sunile Manjee for your reply.I want to use hive ,hbase,oozie,sqoop,falcon,knox,ranger,spark.

Yes,HA required for all services.

monthly 1 TB of data and growing at 10%.

We are aiming for 1 PB data .

Can you brief me little about mutation figure?