Hello, I am a bit confused here. I am currently running an several servers each with an instance of apache 2 and squid 2.5 using: httpd_accel_single_host On httpd_accel_host 127.0.0.1 httpd_accel_port 80 >From your email it seems like I can set the apache servers as cache peers? Is that correct. I ask because I would like to put squid on separate servers but can't figure out a very good mechanism for it it to use all the "origin" servers and detect if an origin server is down. Jerry On Fri, 2006-04-07 at 23:45 +0200, Henrik Nordstrom wrote: > fre 2006-04-07 klockan 09:53 -0700 skrev Ben Drees: > > I am curious about a related question: > > > > In reverse proxy scenarios, what are the options for load balancing > > cache misses among several origin server replicas? > > > I would make each origin a peer to Squid, allowing Squid to use > round-robin etc for balancing the load.. Just remember to set the > connect timeout fairly short for better failover performance. > > > 2) It is my understanding that if DNS returns more than one address for > > a hostname, Squid can be configured to perform round-robin selection of > > an origin server. Are there any caveats to be aware of when persistent > > connections are used between squid and the origin servers? > > Yes, and it does by default. The problem is if/when there is a problem > with one server, it may then become very hard to make Squid use that > server again.. > > > 3) It seems like the Redirector API could be used as a hook to do this > > kind of load balancing also, offering a convenient place to code custom > > health checks. > > I would avoid this if possible. Redirectors is quite resource demanding, > and it's tricky to get the setup correct when rewriting the URLs while > proxied... (cache gets split, backends may get confused about the actual > requested domain etc..) > > Regards > Henrik > > -- Jerry Thomas thomasj@xxxxxxxxxxxxxx I.T. Director ph. 305-662-5959 ext. 242 The SCORE Group fax 305-662-8922
