> >> At least I can point to a full disabling list. Look for test-suite/build > >> tests/layer-01-minimal.opts. Thats everything that can safely be turned > >> off. Though you may find some of them are wanted. > > > > That is what I am looking for! Unfortunately, there is no such list or reference to it in any file in: > > > > 3.1.5 > > 3.HEAD > > > > Oops, sorry I keep forgetting those are not bundled. > You can find a copy of them here: > > http://build.squid-cache.org/job/3.HEAD-amd64-CentOS-5.3/ws/test-suite/buildtests/ Thank you! Exactly what I was looking for! > They build in code that hard-codes the particular file formats. Slightly > faster to log with and less RAM used. My plans are to replace the > options with built-in formats like "squid" and "common" are already. > That will means some of their code is always-present in future, but for > now it can be built without. > > > > >> underscores can be turned off. It enables a DNS violation which makes > >> Squid accept and lookup invalid domains like broken_host.example.com > > > > People want that. They say, your cache is not working (when accessing these hosts). Better to keep it on than dealing with them. > > > > It only works internally because the public DNS does not support them. I > always wondered what types of networks those people ran. What do you mean it works internally? Users browse websites totally unrelated to us. BIND has a check-names option that would allow underscores. Everytime we turned this off, users complained that cache is broken. I must admit I haven't followed this up quite a while. But we keep it on. > > Is this because of an error on my configuration? > > I don't think so. Something else really, really weird is going on. > > Those log lines are built from an array of strings using a code number > for each piece fo text. The CD_* entries are not even built unless you > have "#define USE_CACHE_DIGESTS 1" in your autoconf.h > > My thoughts is perhapse something happend to build an old copy of some > objects. See if a completely clean build still has it. All my builds are clean builds, they run from a script. squid directory is removed, tarball unzipped and configure ran. > > Behaviour is how I want. Only this log entries bothering me. Over the years, I came to greatly trust squid logs. They don't lie. If they point out an error, it is because of me... or Microsoft. > > > > On my main question would the hit go direct with this config if parent2 or parent3 down? > > Depends on the results of random when it was tested to find the peers. > > To find FIRST_UP parent random is run once for each by your > cache_peer_access allow rule. > > * if the first 'fails' its 1/3 chance but second passes. the second is > used, and direct is a backup method. > > * if first passes, the second is still tested and still has 1/3 chance > of being next on the backup list if first is down. direct as a > guaranteed backup again. > > * if both fail their 1/3 tests backup is used. Thank you. J
