Hi. I'm seeing periodic odd behavior from one of our squid2.6 stable18
(and stable22) boxes during the peak hours when the squid is busiest,
but not off-peak, and no other signs of a capacity limit except the
occasional queue congestion warning. About 15% of the requests
for one url we're externally monitoring (presumedly others) come
back with the following headers:
Content-Length: 140493
Content-Type: text/html; charset=utf-8
Age: 1
X-Cache: HIT from oak-tp-squid008
but it only receives 33075 bytes of the payload,
which really is 140493, then the client times out.
In the squid access log:
2008/10/24-14:37:06-0700(PDT) 45626 10.17.2.26 TCP_MISS/200 33075 GET ....
(logformat foo %{%Y/%m/%d-%H:%M:%S%z(%Z)}tl %6tr %>a %Ss/%03Hs %<st %rm)
I am certain that log msg corresponds to that request,
because of unique ids in a custom header.
We are using collapsed_forwarding here. I haven't tried disabling it yet.
Unfortunately, since the problem appears to be load-related, I've been
unable to reproduce for a tcpdump or running squid in debug thus far.
Any ideas how this could be happening?
-neil
