I was working on connecting an AWS EMR to an AWS Managed Elastic cluster. I'd noticed that I could connect to the managed cluster using a curl, but that the curl could not use a port number. Putting either :80 or :9200 would work, the curl request would just hang. As EMR uses 9200 for a default, this also seemed to prevent the EMR from hitting the cluster - a request to it (which worked when sent to our own elastic cluster using 9200) would just hang when configured to hit the managed cluster, whether setting es.port to 80, 8080 or 9200.
I happened to try setting the es.port to -1, and this (apparently) worked as the EMR job completed successfully and the managed cluster did show the indices. So I'm able to proceed with using managed clusters if I like.
However, this behavior isn't documented, which makes me concerned that it may change.
I'm interested whether this behavior was accidental, and/or whether it could be documented and hence become 'official'.