On Thu, 21 Oct 2010, Jeremy Austin wrote:
> Can you give me more details on this? I am switching to 3.x, but if
> it's not ready...
It's pretty stable, there's just some things I had to do explicitly that
were just there in 2.x. The looping issue caused by transparent
interception of HTTP 1.0 queries. A request would be made by some broken
software and becuase of DNAT that request would be translated to a local
request to squid itself... which would ask it self to fulfill the request...
which would cause it to ask itself again... and again...
The loop would eventually eat up all available file descriptors and wedge
the service. In the tests I did after that it seemed that 2.x's loop
detection handled it just fine and killed it. 3.x, however, logged the
detected loop, but let it run rampant. I had to put in ACLs to explicitly
deny any requests made directly to each squid boxes' IP address.
> Not sure how the Flash redir will work with a cluster. Shared storage
> might work. Then again, you could run it separately on each host and
> use rsync to keep a 'shared' cache.
The redirector can redirect to an external web server which would get the
request, cache & fulfill it. You don't have to redirect to the local squid
nodes themselves. If scaling became a problem then maybe a couple of web
servers running off of a shared GFS2 or OCFS filesystem.
--Arthur Corliss
Live Free or Die
---------
To unsubscribe, send email to <aklug-request@aklug.org>
with 'unsubscribe' in the message body.
Received on Thu Oct 21 09:42:50 2010
This archive was generated by hypermail 2.1.8 : Thu Oct 21 2010 - 09:42:50 AKDT