Recently, I have managed to enable HA for HDFS and YARN. Now I have an active and standby namenodes and automatic failover is working properly. I am using Cloudera Manager and CDH 5.
I have a following question.
For example, if my active Namenode crashes and standby Namenode becomes active, is it possible to automatically set up the previously crashed Namenode to promote itself to active when it becomes healthy? That way, the Namenode configured as a standby at the very beginning would simply work as a substitute in critical situations.
Looking forward to your answers!
Have a look at this post to understand failover process:
How does Hadoop Namenode failover process works?
In real life, you don't need crashed namenode to promote itself to active name node again. Let Active Namenode and Stand-by Namenode switch their roles depending on who acquire the lock at ZooKeeper.
It should be immaterial as long as one Namenode is working as active and other Namenode as stand-by and system is highly available.