nothing special for us about it to normal Postgres, works very well.
Kong caches everything to memory so during load it doesn’t hit DB during normal operations. We run about 8kRPM on a t2.medium at 20% load. Most of the APIs use the ACL and JWT plugin.
The only trouble we’ve run into so far isn’t related to RDS itself. When you scale more kong web server instances, it probably needs to fetch a lot of data live from db and acquires many connections to the database. If multiple web instances do that at once, we used to hit max_connections on the db if we weren’t careful.