We are trying to set up a Hadoop automatic fail-over cluster. We are having many issues. the NameNode and Zookeeper FailOver Controller crashes unexpectedly. With all the crashes Both NameNodes are in a active state which will cause a split brain issue with the cluster.
It is simple not working and does not appear to be a robust solution. Please see this URL
For the details of the problem.
We would like to down grade from the automatic fail-over cluster to a manual fail-over cluster. Is there a way to do that. We want to manually tell the NameNodes what state they should be in Active or Standby - is there a way to do this.
We are using the following versions: - hadoop-2.7.3 - zookeeper-3.4.10
You can configure the fencing method to avoid Split-brain scenario,
Check the link and configure HDFS for fencing method.