[squid-users] Debugging intermittent 502 Bad Gateway errors

Hello,

I have a problem that crops up with a single site (alas, https
protocol). It's a subscription service to request financial information
on companies. It works as follows: you add a series of filters to trim
down the number of results to a suitable number: 50, 100, 400 results,
whatever. This poses no problems. Once you're satisfied with the
results, you can then either view the results in full, or ask for an
export (e.g. to a "popular" spreadsheet).

 From time to time, requesting either of these things will result in a
lengthy delay, followed by a "Connection reset by peer" message and a
502 error logged by squid. But it usually works. I learnt from the tech
support of the site in question is that such a request is being proxied
from a back-end server.

After such a failure, a client refresh straight after will bring up the
page correctly. I upgraded to STABLE14 yesterday, and I haven't been
able to provoke the error since. But as it's intermittent, it's really
hard to say.

Anyway, I think the problem isn't with Squid, I think it lies with the
site. At the most it's some weird interaction between Squid and this
site. But I need more evidence to argue my case.

My question is this: what magic debug string should I use to get some
useful information into cache.log? I've tried random sections and depths
by grepping the source, but I either get reams of useless information,
or nothing.

My hypothesis is that their back-end server is getting overloaded, and
is sending some sort of private keep-alive message to the front-end, and
that's leaking out to my proxy, which can't make sense of it. I want to
log more debug information about strange responses, basically. Any tips
would be gratefully received.

Thanks,
David

-- 
"It's overkill of course, but you can never have too much overkill."