Check for CONNECTION_TUNING_PARAMETERS.MAX_SESSIONS_PER_CONNECTION in the broker's Advanced Properties: go to the broker's properties in the SMC's Config view, select the 'Advanced' tab, then click the 'Edit...' button.
By default there's no limit on the number of sessions per connection. I suspect the limit was set on your broker either as a preventative measure against potential rogue clients, or to help debug/diagnose a client that was leaking sessions or using an unexpectedly large number of sessions.
Thanks, Paul.
A further comment on this. A couple of hundred sessions on a connection would be pretty unusual in a standalone JMS application, but it's not particularly uncommon for an ESB container and shouldn't cause any problems.
Say you've got 15 services deployed in an ESB Container each configured with an entry endpoint and 10 listeners. This results in 150 sessions, plus there are additional pools of sessions for sends (see the ESB Connection config).
If you have services that you know are only ever invoked via intracontainer messaging you can omit the entry endpoint for those services (still leaving an appropriate listener count set) and thereby cut down somewhat on JMS resources.
I'm not aware of any known figures for optimum numbers of sessions per connection. There are ESB containers out there running with 1,000+ sessions. By that point I'd tend to be asking whether all those sessions are really needed, and I suspect a good chunk of them are not actively or heavily used, but they're there and not causing any obvious issues.
The message volume and number of parallel message flows is probably more relevant than the number of sessions as such. Messaging on the various sessions is ultimately multiplexed through a single socket if the sessions are all using the same connection, so at some point it may be beneficial to move to a handful of connections, but I personally haven't done any performance testing in this area. Others may be able to chip in here, but if you're concerned about performance you may need to experiment a little with your specific test scenarios.
Sorry I can't provide anything more concrete just now.
Thanks, Paul.
Thanks a lot Paul