On 4/18/06, James Strachan <[EMAIL PROTECTED]> wrote: > On 4/18/06, Hiram Chirino <[EMAIL PROTECTED]> wrote: > > I definitely think we need to implement a process pool type thing > > where we can configure min/max > > processes that we start up per destination etc. Perhaps we do this in > > the broker. Since he is keeping track of how full the consumer > > prefetch is, he's probably the best guy to know when a new consumer > > process needs to get spun up. > > We could just use Jencks which does the pooling of POJOs and threads; > then make each onMessage() call wait until the process quits; then we > can use the thread & POJO pooling to control how much concurrency we > allow per thread etc? >
That would be possible but then Jencks need to deliver the message to the spawned process since it has consumed the message from activemq. So Jencks would end up being something like a CGI container almost like an httpd server. Which is cool but different from what "runmqchi" does. runmqchi monitors queue depths and launches processes to consume those messages. Those processes connect back to the message broker to actually get the message. > -- > > James > ------- > http://radio.weblogs.com/0112098/ > -- Regards, Hiram
