Euler Taveira de Oliveira <euler(at)timbira(dot)com> wrote:
> Talking about your problem, are you sure you're not reaching
> max_connections?
It also strikes me that from the minimal information given, it might
be possible that pid numbers or port numbers are wrapping around
before the OS is ready to allow re-use. I haven't seen that
behavior except in machines infected with a worm, but this test
might be edging into the same territory if it's using a new
connection for each request. Obviously, nobody who cared about
performance would use that technique in production, but it rather
sounds like this test does.
-Kevin