On Thursday, April 12, 2001, at 10:34 PM, Graham Leggett wrote:
Chuck Murcko wrote:
[Thu Apr 12 12:19:54 2001] [error] (32)Broken pipe: core_output_filter:
writing data to the network
Did the browser break the connection here? If you get a lot of these
please tell me - I think we might be writing too much data (the browser
might be saying - oooh, that's "content-length" bytes, better close the
connection now, when there is more on the way)...
I see clusters of these, like from every server connected to the browser
in the (prefork) pool at once, and the browser is breaking the
connection(s).
[Thu Apr 12 12:20:04 2001] [error] [client 10.0.1.4] proxy: error
reading status line from remote server www.channelcincinnati.com
[Thu Apr 12 12:20:05 2001] [error] (111)Connection refused: proxy:
attempt to connect to 208.45.172.82:80 (scrooge.ibsys.com) failed
The last line is normal; it's the first few that I'm looking into.
Do we go somewhere bad if we don't read a good status line from the
origin server? All servers in the pool are running against
www.channelcincinnati.com in this timeframe (site's got a lot of
graphics).
As I understand it we HTTP_BAD_GATEWAY if the backend server returns
anything dodgy.
Yes, and this execution path looks safe.
Chuck Murcko
Topsail Group
http://www.topsail.org/
