At 11:21 AM 4/12/2002 +0200, Ulric Eriksson wrote:
>Not really. It looks like a bug. Can you reproduce it reliably?
I have not been able to consistently reproduce it. It is not bound to any
type of content (cgi scripts, static pages, or dynamic pages), nor does it
seem to be tied to any particular server.
>I suggest using the -n option since the nonblocking stuff only seems to
>slow things down. In the benchmark I referenced in an earlier mail,
>running http_load with the -checksum option didn't produce any such
>errors.
Using the nonblocking option actually added a new problem: after several
connects, my requests wouldn't get forwarded and the browser would hang.
http_load with the parameters lower (-parallel 5) than the ones you used
actually caused FreeBSD's rate limiter to kick in and start dropping
connections.
This archive was generated by hypermail 2.1.2 : Fri Apr 12 2002 - 16:06:57 CEST