CODEWITHSUNDEEP
hbaseapache-zookeeper
Jack
Jack

Reputation: 5880

How many votes Zookeeper consensus do in Hbase

This maybe a simple silly question, I only know keep Zookeepers odd number for master voting consensus. Do they also vote for other things in Hbase or just for selecting new master? For example: do they vote for Region load balance or Root Region server? thanks!

Upvotes: 0

Views: 198

Answers (1)

halil
halil

Reputation: 1812

I am not sure what you asked but i don't think a hbase slave node ( region server) can be a hbase master node if master node dies. They are separate applications as far as i know. So zookeper can not play a role for this.

Edit:

Zookeper does not vote for other applications, only votes for selecting zookeeper master and you should have 1-3-5-7 zookeeper nodes based on work load on zookeeper usage, lots of apps use zookeeper to exchange data between master and slave nodes, and zookeeper is a disk io oriented application. If you have 5+ hbase nodes you should have at least 3 zookeper nodes to be safe, i think.

Hbase master backup mechanism is having multiple master nodes, one of them is active master and others backups.

You can read more here :

http://www.cloudera.com/documentation/enterprise/5-2-x/topics/cdh_hag_hbase_config.html http://blog.zahoor.in/2012/08/hbase-hmaster-architecture/

Active Master Manager

Handles everything on master side related to master election. This is the place where the backup masters block, until the active master fails or the cluster shuts down. Listens and responds to ZooKeeper notifications on the master znode, both nodeCreated and nodeDeleted. Uses a zNode called “master” under the base zNode.

Upvotes: 1

Related Questions