I have a question about routing. We are using ES for real-time analytics. We are routing documents based on a unique customer id field. We just noticed a big WARNING at the top of an article about routing that says we should no longer follow the advice in the article - https://www.elastic.co/blog/customizing-your-document-routing.
I'm trying to find some more details on which part of the article is now considered bad advice. Should we be using routing to ensure all data for a particular customer ends up on the same shard?
I think the reason for the warning is that we saw many users abusing routing and having issues like unbalanced shards because some users have many more docs than others and bottlenecks on one or two shards because the most active users in terms of request throughput are on the same shard. Routing can be useful but should be used wisely, and in particular you should always be ready to move some users to a dedicated index in case they become "too big" for this multi-tenant setup.
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.