Henrik Nordstrom wrote:
On Sun, 30 Jan 2005, Jim Richey wrote:
Running Squid 2.5STABLE7 on Slackware 10 Linux with kernel 2.6.10 and Secure Computing SmartFilter patch.
Please verify that the situation is the same without the SmartFilter patch. In the response below I am assuming you have already done this.
For many reasons I can not support proprietarily patched versions of Squid. (legal, political and etical reasons, all applies).
Ok. I'll have to get the standard release loaded on some test systems and see if I can duplicate the problem.
Squid is configured with:
./configure --enable-async-io --with-aio --enable-auth=ntlm,basic --enable-removal-policies --enable-cache-digests --enable-kill-parent-hack --disable-ident-lookups --enable-external-acl-helpers=wbinfo_group --enable-ntlm-fail-open --enable-smartfilter
squid-2.5.STABLE7-response_splitting.patch appears to break the retrieval of cache digests from peer caches. After implementing this patch I get the following in the cache.log:
2005/01/30 13:03:54| temporary disabling (Not Found) digest from proxy4.highmark.com
and in the access log I get:
1107071002.214 1 167.164.1.11 TCP_MISS/404 246 GET http://proxy2.highmark.com:9119/squid-internal-periodic/store_digest - NONE/- text/plain ALLOW "-"
Have you waited for the digest to be recomputed? When you restart Squid it takes a while before the digest has been recomputed.
debug_options ALL,1 71,2
will tell you more about the digest generation process.
Any errors in cache.log on the digest providing peer?
After an hour I still get the (Not Found) message in cache.log. I don't get any other errors. Store digest in cache manager shows that a digest has been created. However, on the peer, in peer selection algorithms, the peer digest state shows:
needed: yes, usable: no, requested: no
Regards Henrik