https://bugzilla.novell.com/show_bug.cgi?id=474217
https://bugzilla.novell.com/show_bug.cgi?id=474217#c9 --- Comment #9 from Lluis Sanchez <[email protected]> 2010-12-14 12:22:49 UTC --- When I implemented the Tcp channel I did not use the system thread pool because I had bad experiences with it. For example, let's say you have a remote method which directly or indirectly uses the thread pool. Let's also say you have a limit of 50 thread pool threads, and then you make 50 concurrent calls to that remote method, which are dispatched using the thread pool. This will result in a deadlock of the whole application, since the 50 threads are waiting for one more thread to be available, which will never happen. Maybe this is an extreme case, but I remember hitting this issue in the old days when doing some stress tests. I just built a test case that reproduces the deadlock by constraining the number of available pool threads, so the issue is still there (I could reproduce it with the http channel, which uses the threadpool, not with tcp). I don't know how likely is to hit this issue nowadays in real world applications. The thread pool limit is not higher, so maybe it is not so likely. However I still feel uncomfortable about using the system thread pool as a request processing pool, but if you guys think that there is no real problem with it, I'm ok with doing the change. In any case, the PoolGrowDelay issue is a bug that can be fixed. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are the QA contact for the bug. _______________________________________________ mono-bugs maillist - [email protected] http://lists.ximian.com/mailman/listinfo/mono-bugs
