Hello,
I face a huge drop in performance when I work with huge arrays (thousand elements).
Is it something expected ? Is there a way to understand how arrays are internally handled by elasticsearch ?
What kind of performance are we talking about? Indexing or searching? Can you quantify the slow down?
I check my benchmark after your question and I have to apologize because it was my mistake. A part of my application was not well tested. So there is no more drop of performance.
But still, I wonder if it's a good practice to have big arrays in an elasticsearch document ?
It depends how large it is. In general, for in field that is only indexed, things should be fine, the index is built to handle high-cadinality fields gracefully. However I would be more careful with doc values which tend to like better when documents have a reasonable number of values (say a couple dozens at most).