I presently come with an application used using Tomcat that interacts having a Postgres database via JDBC. The queries are extremely costly, what exactly I am seeing is really a timeout triggered by Tomcat or Apache (Apache sits before Tomcat during my configuration). I am attempting to limit the connections towards the database to twenty-30 synchronised connections, to ensure that the database isn't overcome. I have carried this out while using .. configuration, setting maxActive to 30 and maxIdle to twenty. I additionally knocked in the maxWait.
Within this scenario I am restricting using the database, however i want the connections/demands to become Put within Tomcat. Apache can accept 250 synchronised demands. So I have to ensure Tomcat may also accept this many, but handle them properly.
Tomcat has two configurations within the HTTP Connector config file:
- maxThreads - "Max quantity of request processing threads to become produced through the Http Connector, which therefore determines the max quantity of synchronised demands that may be handled."
- acceptCount - "The utmost queue length for incoming connection demands when all possible request processing threads are being used. Any demands received once the queue is full is going to be declined."
So I am speculating when I set maxThreads towards the max quantity of JDBC connections (30), i quickly can set acceptCount to 250-30 = 220.
I do not quite comprehend the distinction between a thread that's WAITING on the JDBC link with open in the pool, versus a thread that's queued... My thought is the fact that a queued thread is consuming less cycles whereas a running thread, waiting around the JDBC pool, is going to be investing cycles checking the pool for any free thread...?