> We have 2 Squids (on server27 and server28) in front of one webserver > (server18), which is not running Squid. Any time 27 or 28 have a MISS > on a request they immediately query FIRST_PARENT_UP, which is always > server18. Even when we change it so that 28 is a parent of 27 and vice > versa. In other words, the caches don't pull cache from each other, > even when we set it up that one HAS the object the other needs; it still > goes to the PARENT, server18 for the file. We can't figure out why this > is happening. So, we have the following layout: > > INTERNET > / \ > / \ > V V > Server27 server28 > Squid squid > | | > | | > V V > web server/backend/originserver > no squid running on this box > > Relevant config lines, these from server28: > ------------------------------------------- > cache_peer server27 sibling 3128 3130 proxy-only login=PASS > acl fromneighbor src 192.168.120.225 # IP for server27 > cache_peer_access server27 allow fromneighbor > cache_peer server18 parent 80 0 no-query no-netdb-exchange no-digest > originserver login=PASS > > Does anything look wrong? Why wouldn't server27 and 28 talk to each > other and use each others' cache? Squid prefers the fastest source. You could try the weight option to shift the balance toward the sibling. As a side note: "proxy-only" prevents locally storing objects received from the sibling. Its a policy choice, but you may want to increase your HIT rate by allowing those to be cached. Amos
