Support Questions

Find answers, ask questions, and share your expertise
Announcements
Celebrating as our community reaches 100,000 members! Thank you!

Need suggestion to stop or Decommission Datanode?

avatar
Explorer

Hi , We are using 12 Datanodes in our cluster. And we have 3 replication factor.

Currently we are doing cost optimization and our total cluster is 35% used. We want to remove one datanode from our cluster, but we want to see first how our cluster performs on 11 datanodes. if it doesn't perform well, we need to restart that datanode again.

To experiment this- Do I need to Stop the datanode or Decomission datanode?

Please need suggestion.

1 ACCEPTED SOLUTION

avatar
Master Mentor

@Madhura Mhatre

It's well documented in by hortonworks Once you launch the decommissioning the blocks on that node will be distributed to the other node remaining nodes

108134-decom-datanode.png

If the replication factor is higher than the number of existing data nodes after the removal, the removal process is not going to succeed!

View solution in original post

4 REPLIES 4

avatar
Master Mentor

@Madhura Mhatre

It's well documented in by hortonworks Once you launch the decommissioning the blocks on that node will be distributed to the other node remaining nodes

108134-decom-datanode.png

If the replication factor is higher than the number of existing data nodes after the removal, the removal process is not going to succeed!

avatar
Explorer

@Geoffrey Shelton Okot Thank you for your response. Instead of Decommissioning the node Can I stop it?

What is the difference? Whats the best way to stop it or Decommission?

avatar
Master Mentor

@Madhura Mhatre

Surely you could just t^do that but what happens to the replicas stored on that particular data node? Somehow your cluster has to reconstruct those replicas if you had a replicator factor of more than 1.


Here I was talking about planned maintenance! Just switching it off will also force your cluster to do the same thing in the background with alerts and ONLY when the replicas have been reconstructed will those alerts go away.
There is a performance cost for both decommissioning and just unplugging the data node.

avatar
Explorer

@Geoffrey Shelton Okot Thank you. I think Decommissioning is a best way to do this. Do I need to turn on maintainance mode while I do decommissioning datanode? Also Do I need to stop the services like - NodeManager, Ambari metrics for that datanode after decommissioning?