I just upgraded to ELK 5.4.1 and was looking and noticed that my Primary nodes are not necessarily on the 4 nodes. Now I am sure it does not affect search as any node can be used.
But I do index a large volume of data every day, ~1TB ( ~20k messages a second)
This does not affect indexing or searching since primaries and replicas need to perform the same amount of work in both cases. So Elasticsearch does not try to spread primaries evenly across nodes.
There is only one exception, which is the update API, which always runs the script on the primary and then replicates the updated document to the replica, so the replica does not have to perform the update again, which might trigger an uneven load. But other than that you should not worry. And there are plans to address the update issue: https://github.com/elastic/elasticsearch/issues/8369
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.