I'm running into a problem under high client (not server) load. If I create a client process that makes only light requests to the server, everything works perfectly. If, however, I spawn a bunch of client requests simultaneously, using HTTPDownloader and HTTPClientFactory (or getPage and downloadPage, it doesn't seem to matter), I invariably get a bunch of "TimeoutError: User timeout caused connection failure" failures. These failures have no stack associated with them, so it isn't clear where they are originating, but it appears that they are due to timeouts on waiting for the connection setup, not on the response.
I know that there is a lot of activity on redoing the client in web2, but I didn't see any mention of this particular problem (or potential ways to avoid it) in the list archives. Is this a known problem? Is there a workaround I can rely on until something better comes along in web2?
The breakpoint (on my 2.8Ghz Pentium) seems to be around 40-50 simultaneous client connections. On my slower 2.53 Ghz machine, it is closer to 70-80.
(I can produce a short example that reproduces the behavior if needed)
__________________________________________________ Do You Yahoo!? Tired of spam? Yahoo! Mail has the best spam protection around http://mail.yahoo.com