I have a situation where we want to upgrade hardware of datanodes.
I know that I can decommission each of the machine and rejoin them.
However, I was looking for ways to avoid replication of terabytes of data to another nodes as this is planned downtime.
I came across an option where I can increase heartbeat recheck interval (dfs.namenode.heartbeat.recheck-interval) to delay the time namenode take to identify a dead datanode.
To use this option I am sure I need to restart namenode deamon.
Is this the way cloudera recommends or is there any other ?
Thank you all in advance.