I got all my training back under 3.5 in the stone ages. At the time I got a good feel for the way messages passed through the engine and what the recovery database states corresponded to in the engine.
In the near future we’re going to be adding multiple new facilities to our network and we’re trying to be proactive about understanding the load on our lock managers today in a large multi-site implementation and predict how that load will change as we add the new facilities so we can reorganize our sites and processes accordingly before we run into problems.
To that end I have pulled out my (current) Reference Guide and gone digging. I have a good definition of the Engine Internals -> Engine Components -> Queues. I also have the table defining the Recovery Database States. What I am trying to get (and may have missed in existing documentation somewhere) is a list of the actual points where a message with “Use Recovery” set will write to the recovery database during normal processing (e.g. I know either Q4 or Q5 or possibly both cannot be written to the database under normal processing because there aren’t states available for both).
I’m trying to reconcile my recollections that back under 3.X a message which processed normally without interruption or error didn’t write to the database with states 2, 4, 8, 9 or 10 with the current documentation and the addition of state 16 for the new recovery model.
If there’s existing documentation or a thread that my search-fu failed to find, a link to that or a heirarchical location in the documentation would be appreciated (and sufficient).