Hey Rob, we run into this problem too with one of our applications.
One application sends several million messages a day, which are processed promptly. For some other queues we have a steady but low rate of messages and intermittent clients which connect occasionally. So we end up with a small (~1K) number of outstanding messages but millions of messages (mostly acked) in the store until those clients connect and pick up their backlog of messages. I don't know if we need such finegrained resolution as a store per queue since this leads back into the issues with too many file descriptors that were apparent in the AMQ Store, and is resolved with KahaDB. It'd be nice to be able to define a set of stores, with one as default, and destination policies to direct specific destinations to specific stores when needed. I know we looked at this at one point with Gary, but I don't know how much progress we made. Perhaps your code might be a good starting point for us to pick this up again. cheers, James. On 15 September 2010 19:13, Rob Davies <rajdav...@gmail.com> wrote: > I did this a while ago too see if it would improve performance - it didn't so > I ditched the code. Why do you want to do this? > On 15 Sep 2010, at 18:02, nervousbadger wrote: > >> >> We've decided that we will need to implement per-queue log files. >> >> Do you know if anyone has attempted this before? Are there any code >> templates we could look at that would give us a bit of a kickstart? >> >> Thanks, >> Pete. >> -- >> View this message in context: >> http://activemq.2283324.n4.nabble.com/KahaDB-log-f-tp2539334p2540873.html >> Sent from the ActiveMQ - User mailing list archive at Nabble.com. > >
