Question-91: What is the issue with the deleting the slave components?

Answer: Deleting a slave components like DataNode (Slave of NameNode) does not automatically inform a master a master component such as a NameNode, to remove it from its exclusion list.

And if you want to add back that deleted component into the cluster then this added back component remains decommissioned from the master perspective’s and to solve this problem we have to restart the master components.

If you are deleting the host than make sure you stop all the components on that host. And then if DataNode is running on that host then stop it first.

Question-92: Suppose you are having single master cluster and want to delete that host, what is the resolution?

Answer: You must follow the guidelines provided which is controlled using the guardrails to prevent any accidental deletion of the node/hosts which is the only master in the cluster. And before deleting such hosts, we need to first move all the master to a different host.  For guardrails are like this if you want to while deleting the NameNode, it checks the minimum number of the required components. If it is going to be less than 1 after the deleting then it would not delete the NameNode, because it is for sure accidental request.

Question-93: Why do you want to use Maintenance mode for a component?

Answer: Using the maintenance mode we can suppress the alerts when performing hardware or software maintenance, or while changing the configuration settings etc.  And also on that service, component or host then no bulk operation can be performed, if it is in maintenance mode. And you have to explicitly start and stop the each individual service.

Question-94: Why do you want to use the maintenance mode at the first place?

Answer: There are some common instances for which you want to run the service/host/component in maintenance mode like testing the configuration changes, to delete a particular service completely, and suppress the alerts. So there are following requirement which can be taken care with the maintenance mode

  • Prevent alerts generated by all components on this host.
  • We can have capability to start, stop and restart each component on the host.
  • Prevent host-level or service-level bulk operations.

And if we put the host on the maintenance mode then all the components on that hosts are implicitly go in the maintenance mode.

Question-95: Suppose you put one of the DataNode on the maintenance mode then what all you can think of get affected?

Answer: If we put one of the DataNode from Hadoop cluster following things can happen to services if they are running.

  • All the alerts on the DataNode would be suppressed.
  • This particular node will be skipped from HDFS start/storp/restart and any Rolling start.
  • This node would be skipped from any bulk operations.
  • This node will be skipped from the host-level restart operations.