We are using the software appliance version, with remote database. And never in my life did I expect that our environment would experience network problems.
But Murphy caught up with me, and now some days we get tens of thousands "connection lost" in our logs.
The problem IS NOT in the database.
Our Oracle cluster is barely showing any load, and the SCAN-listeners are ready to move us IF there would to be a database issue.
Instead as said, we have a congested network, and it is primarly any of the 7 network pools in Wildfly that is loosing connections, that in turn is causing "transaction problems" (although my knowledge exactly how wildfly is handling a connection lost while doing a transaction is sparse, and even worse when knowing how Aveksa responds to this [if at all])
Also it seems that there are alternative routes to the database, if using secondary clusters of the database. But the "Oracle SCAN service" will never bother with that if "the database workload is low".
So I tried to manually use a more advanced jdbc-string in our test Environment, but I never managed to get Aveksa to connect to both remote database (physical) locations. (although not an expert in the wonders you could do in a jdbc expression, AND it also seems that sometimes "wildfly flags" are used in Wildfly, which I do not know anything about)
Anyone had the same issue (congested network) with this product, and has some advice to give?