On Sat, 17 Jun 2000, Hamid Reza Shahriari wrote:
> I could get 300/sec with Poly-2.2.9 no-proxy test.
> But I got some errors by polyclt:
>
> 002.06| i-warm 2914 47.91 2477 0.00 0 400
> 002.06| Connection.cc:172: error: 1/1 (104) Connection reset by peer
> 002.06| Connection.cc:172: error: 2/2 (104) Connection reset by peer
> 002.06| Connection.cc:172: error: 3/3 (104) Connection reset by peer
> 002.06| Connection.cc:172: error: 7/7 (104) Connection reset by peer
> 002.06| Connection.cc:197: error: 1/8 (11) Resource temporarily
> unavailable
> or
> 006.65| i-inc1 11784 42.00 2503 0.00 0 364
> 006.70| Connection.cc:172: error: 63/80 (104) Connection reset by peer
> 006.71| Connection.cc:172: error: 1/95 (32) Broken pipe
> 006.73| i-inc1 12091 72.91 2519 0.00 59 396
> 006.75| Connection.cc:172: error: 2/139 (32) Broken pipe
> 006.78| Connection.cc:197: error: 31/141 (11) Resource temporarily
> unavailable
> 006.81| Connection.cc:172: error: 127/162 (104) Connection reset by peer
>
> I have 128M RAM in both client/server, Max fds is 1024. Is this
> efficient?
You will probably need more FDs when you run through a proxy or increase
request rate. 12K FDs should be sufficient for most purposes. Most
errors you are seeing are probably due to HTTP race conditions and,
hence, are "normal". Future Polygraph versions will handle those race
conditions more gracefully. The "Resource temporarily unavailable" error
is of a different kind. It is probably caused by a minor Polygraph bug
that is fixed in the upcoming version.
Note that I can get at least 600 req/sec on a no-proxy test through a
switch, on two (or more) 450 MHz P-IIs running FreeBSD.
Alex.
This archive was generated by hypermail 2b29 : Tue Jul 10 2001 - 12:00:13 MDT