Hi all, I think I‘ve finally managed to spare some time to finish my work on the connection cache.
I think https://github.com/apache/plc4x/pull/818 tried to address a problem, where if something happens with the connection while it’s being used, that it might be in a inconsistent state. So, it was the goal to intercept any errors occurring during the execution of any request and to instantly close the connection if anything went wrong. However, I wasn’t happy with the PR as it sort of brought the connection into an odd state and it only was a partial solution. Also did it change the API, weakening the contract between PLC4X and the using application. And if there were requests waiting for the connection, they would just have been dropped and would never get a connection. So, what I did, was something similar to the original PR: I intercept the execution of all types of requests (not only read- and write-requests). But instead of closing the connection instantly, I mark it as “invalidateConnection” … when the close() method is called, this tells the connection container to invalidate the connection. The connection container then handles closing the connection and possibly creating a new connection and using that to serve the waiting clients. I hope this solves the problem people were seeing with the current implementation. I will then close the PR818 if I don’t hear anything in the next 72h Chris
