leader_follwer.txt

Kafka stores stream of records in catageries called topics.
Each topic stores data in partitions.

To achieve high availablity,kafka will replicate partitions.each copy of partitions calleds as replicas.
There are two types of replicas in kafka,
Leader Replica and follwer Replica,

Leader Replica : Out of all this replicas ,consumer and producer use a specific replica to produce and consume messages.
		 All read and write request for a particular partition will always served by a specific replica and those replica is
		called Leader Replica.
		Each partition has a single replica desginated as Leader.
		Leader Replica is responsible for reading and writing partiton.
		With out leader replica ,you cannot perform read and write a data to or from a partition.
		Apart from read and write request from producer and consumer,the other responsibility of leader replica is keeping a track of repicas up to date.
			

Follwer Replica : All replicas which are not leader replica is said to be follwer replica,
		 The only job follwer replica is keep update with the leader replica by fetching a message from leader replica..
		At the event of leader replica failure,one of the follwer replica act as leader replica.
	           
--> in order keep himself (follwer replica) update with the leader,follwer always connect for every 500 miliseconds 
				----   replica.fetch.wait.max.ms -----
--> in every fetch request ,follwer sends offset number that he want to fetch from the leader.
--> the rule is follwer replica could not request a random offset fom the leader.
     example in order to request offset 4 ,the follwer must offset number 3
--> if any follwer who is not requesting a new message from the leader within 10 seconds (replica.lag.time.max.ms) considered as out of sync.,
--> replicas who are continously in sync with leader are called as in sync replcas (ISR)