During scaledowns, I want to update the seed_hosts value (to remove the deprovisoing nodes).
I want to apply this new value to all the nodes which are still remaining in the cluster.
Is it required to update the seed_hosts value when a scale down occurs?
No, not required, but it's good practice. Note that the seed_hosts value should only mention the master-eligible nodes, and it's rare to want to scale those down (there's normally always three of them).
DNS is a popular choice: create a DNS entry that resolves to the addresses of your three master nodes and set discovery.seed_hosts to the name of that entry.
Unfortunately, I have a different scenario.
Each node in my cluster is a master as well as a data node.
So a 36 node cluster will have 36 master eligible node and we have to support scale down form 36 ->1 node as well.
I believe in such scenarios, scale down becomes tricky. We need to evacuate these nodes so that quorum doesn't break but I think we should also update the seed_hosts.
Apache, Apache Lucene, Apache Hadoop, Hadoop, HDFS and the yellow elephant
logo are trademarks of the
Apache Software Foundation
in the United States and/or other countries.