CODEWITHSUNDEEP
hadoophbase
Jack
Jack

Reputation: 5880

how to solve the single node failure of hbase namenode

I'm using hbase 0.94.0,I set up three machines,one for master and the other two machines as region server,also the master is namenode server.you know the hadoop namenode is a single node,so I'm very afraid of the single failure,do you have good ideas to solve this problem. Thanks in advance

Upvotes: 1

Views: 3334

Answers (1)

André Staltz
André Staltz

Reputation: 13994

There are two issues to handle here: HDFS's NameNode failure, and HBase's master failure.

For HBase: You can employ multiple HBase masters. Only one of these masters will be active, and the remaining are backups. One of the backups takes the role of the primary once the previous primary fails.

For HDFS: the Namenode is a single point of failure, so try to employ very reliable hardware for it, with RAID storage or some other redundant storage solution. However, there have been some interesting alternatives to backing up the Namenode, such as the Backup Node and AvatarNode

Plus, it is generally recommended to install HBase masters on different nodes than Hadoop Namenodes. I don't know so much about your applications, but your installation is small (3 machines) and doesn't seem to be in production, so you can actually manually restart the master/namenode. Failures are only a serious concern when many machines are being used in a production environment.

Upvotes: 2

Related Questions