On Mon, 3 May 2010 13:28:21 -0700, David Raccah <raccah@xxxxxxxxx> wrote: > Hello, > > Please excuse the newbie. I checked most of the search engines on > squid pages and could not find what I was looking for. Though it may > be because I did not use the correct keywords. > > So we have a large set of squid boxes sitting in front of some slow > running code. The data is mostly static, so we use squid as a proxy > and it caches the data. The TTL on the cache for now is 1 week or > more, and so we are saving the backend/origin from being pounded and > love it!!! However, we are seeing a large number of near-hit instead > of pure hits. For us a near-hit is equal to a miss, because it caches > the cache (L1 and L2) to go to the origin/backend. We are using HTCP > to clear the cache when there is a change (much like wikipedia does), > so we can trust that our L2 is as close to fresh as possible. > > So: > > 1) Since we can guarantee that the L2 will have the latest > information, is there a way to ignore the "if-not-modified" header? > Depends on where it is being generated and exactly which if the If-* header it is. (there is no if-not-modified header). > 2) is there a way to declare the L2 cache as the origin-server instead > of just a parent cache - not a great approach, but need to mitigate > going to the origin if the L2 has a hit? Yes. Setting "originserver" on the parent cache_peer. However I think ICP/HTCP are not sent to origin servers. > > 3) is there a utility to update the timestamp of the cached objects. Maybe the squidpurge tool. I have not yet looked at it closely. Amos
