On Tue, 8 Mar 2005, Allen Schmidt wrote:
We have squid in front of Zope (ZEO with 2 public clients) as caching and accelerator.
Ok.
We have a redirector written in python to bounce between the two clients.
Why a redirector? Squid has balancing already built in.. Admittedly simple, but still quite effective.
Occasionally the cache.log will show:
2005/03/08 14:13:26| Failed to select source for 'http://st.sageanalyst.net/tag-703.js'
2005/03/08 14:13:26| always_direct = 0
2005/03/08 14:13:26| never_direct = 1
2005/03/08 14:13:26| timedout = 0
no live cache_peers where this could be forwarded, or no peers where this request were allowed to be forwarded.
if these sites are not yours then most likely someone tried to use your accelerator as a general purpose proxy. See access.log.
it is recommended to set up http_access to only allow reuqests to your published servers using the dstdomain acl. This will stop these requests proper, avoiding the "failed to select source" clutter in your cache.log.
2005/03/08 14:13:26| WARNING: redirector #2 (FD 7) exited
This is worse.. looks like your redirector is not entirely stable.
but limiting access to your sites only in http_access may help, as this also limits what gets sent to your redirector.
And recently I have seen our main IP address with /robots.txt on the end and the same message.
We have a robots.txt file so why can it not find it??
Most likely your forwarding policies or redirector does not account for requests by IP.
Regards Henrik
