zookeeper-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Ibrahim <i.s.el-san...@newcastle.ac.uk>
Subject Re: Observer write behavior with partition
Date Thu, 02 Oct 2014 16:34:53 GMT

* A client connected to an Observer sends, say, 3 consecutive writes 
* The Observer then loses connection to the leader (network partition)

What is the outcome?

* The writes are queued on the Leader instance, correct?
Yes, you are right. In this case, the leader will trigger the broadcast
protocol with available follower (quorum are needed). 

* Therefore, the writes will succeed, correct?
Yes, you are right. The writes will success if quorum (majority of ZooKeeper
servers) agree upon the write requests.

* When the Observer repairs the connection it will see the 3 writes,
correct?
Yes, you are right. Returned observer will see the 3 writes and any missed
update requests. This achieves when first the observer reconnect to the
ZooKeeper cluster, the leader will send /diff / (any missed writes) to make
the observer up-to-date.


Regards, 

Ibrahim



--
View this message in context: http://zookeeper-user.578899.n2.nabble.com/Observer-write-behavior-with-partition-tp7580368p7580369.html
Sent from the zookeeper-user mailing list archive at Nabble.com.

Mime
View raw message