On Mon, 2006-04-24 at 18:47 -0400, Miguel de Icaza wrote: [...] > As for the second problem, the one I mentioned: we found that this seems > to be caused by the CPU being too busy with the generation of the return > page to have time to close the dangling sockets. > > At least that is what it seems so far (when I used Apache Bench to flood > the machine with requests, which was the only way I would get the reset > event). > > Am not sure how Apache copes with this problem, probably they have a > larger listen queue than we do.
We use the same backlog length as apache, so it must be something else. I'll try to figure out what it is. -Gonzalo _______________________________________________ Mono-list maillist - [email protected] http://lists.ximian.com/mailman/listinfo/mono-list
