On 25/09/2013 6:30 p.m., Luis Daniel Lucio Quiroz wrote:
Hello :)
For a reason, dont ask. I need to configure a Squid that is able to
do reverse lookup from a webapp that does dinamic hostdomains. So i
need to configure a squid that does *.mydomain.
I dont have too much control on *.mydomain IP's, so cache_peer wont be
very helpful in this case. I need to relay in DNS for this. Is it a
way to make this work?
If you use DNS in the reverse-proxy to access the origin servers you
will need even more control of the IPs than you do with cache_peers.
Since both configurations require the public DNS needs to be configured
to point the clients at your proxy. But a reverse-proxy without
cache_peer *also* requires a reliable split-DNS configured because those
DNS servers become responsible for preventing forwarding loops and must
not send Squid the same IPs which are going to the clients.
The configuration in Squid is simple, but leaves you open to a new set
of DNS-related security problems such as DNS spoofing or hijacking in
addition to the regular HTTP security issues.
Starting with the simple virtual-hosting reverse proxy configuration:
http://wiki.squid-cache.org/ConfigExamples/Reverse/VirtualHosting
* You drop the cache_peer directives.
* You replace the cache_peer_access with always_direct.
always_direct should use the same ACL rules as you would have placed
on cache_peer_access.
NP: When defining the dstdomain ACL of hostnames you can use the
wildcard form of dstdomain:
acl outsites dstdomain .mydomain.local
HTH
Amos
