> Squid Users: > > I am trying to determine which set of timeout-related configuration > options would result in the following behavior: > > We currently use a squid proxy as an HTTP accelerator for serving > graphical charts. The hostname 'charts.altosresearch.com' resolves to an > IP address/port on which squid (v3.0) is running. The 'defaultsite' > option is set to 'www.altosresearch.com', which actually generates the > graphical charts. All of this works very well: > > http://charts.altosresearch.com/AltosCharts/CA/SAN+FRANCISCO/94115/median_price/a/1,2,3,4/m/e/mf/552.png > > > However, there are times when in production we have to take the > apache/tomcat server at 'www.altosresearch.com' offline for maintenance. > During these times, any requests to 'charts.altosresearch.com' will fail > because the squid proxy's attempt to validate the cached version fails. You appear to be mistaking the purpose of 'defaultsite'. The only 'defaultsite' in squid is in the http(s)_port settings. There it means the Host: header name to use if the client does not specify one. Nothing more. > > Instead, I would like to configure squid to observe the following rule: > If the connection to www.altosresearch.com fails (returns 400/500 code, > or connection times out after XX seconds), then attempt to serve the > response using a cached object. The configuration variable > 'minimum_expiry_time' seems related, but not sufficient. Then you want to try the Squid 2.7 development release and add some cache-control headers to your servers output. http://plone.org/products/cachefu/roadmap/9 http://www.squid-cache.org/Versions/v2/2.7/ Amos
