Hadoop High Availability with Linux HA

摘要

This paper proposes The High Availability Method using Heartbeat and DRBD on Hadoop Namenode Failover. Implementation of Hadoop as computing process framework and media storage needs to be supported by the availability in case of failure. Cluster system utilization in Hadoop with many Datanode is one of the strength points. Stored file will be replicated in distributed datanode as blocks that makes if one the datanode server fails, the other datanodes server still has duplicates blocks. The problem in hadoop system, there is just one Namenode which is the communicator between users and Datanodes. While Namenode goes down then users can't access files stored in Datanode. This proposed method accomplishes the Namenode failover process. The use of this method is very effective because it can accelerate downtime with automatic failover formed by Heartbeat and real-time mirroring formed by DRBD. The results indicate that the use of DRBD and Heartbeat may accelerate failover time of less than 10 seconds.