Hello :)
I have something to report about grizzly http 1.9.4
Fist of all, sorry for the poor information, but it's quite difficult
to track the problem.
I have configured a grizzlywebserver with many worker thread
(changing the default stat pool minthread)
this problem (i will describe it after this) happen with all the
tested configuration (5,50,100 worker thread)
When grizzly reach more than 60-70 concurrents users, it simply
blocks for 30 seconds from time to time.
I have profiled the code, to see what's up with threads (i have used
jprofiler)
what come out is that worker thread are "free" (waiting) most of time
(even with 5 workers)
while a thread called "grizzly selector"
became green (running) for a while.
In this timeframe, all the request are blocked.
When grizzly selector (i don't have the correct name, i am afraid)
come back to waiting,
all the comunication restart suddently.
There are some trick to avoid this problem?
best regards
Rama