Hi, i'm trying to figure out a good configuration for reverse proxying / load balancing an Apache cluster. The main directives:
* for each apache: cache_peer this.is.the.ip parent 80 0 no-query * set peer_connect_timeout to something like 2 seconds Now the questions: * If the TCP connection times out after 2 seconds, will Squid move on to the next parent without returning an error to the user? (I know about the mark-as-dead process, but afaik this kicks in after a minute or so) * Is there a way to take the load of individual parents (Apaches) into account? I was thinking about something that would delay ICP queries depending on the current load, effectively lowering the machine's priority in the selection process. My understanding of the selection process in the no-query configuration is that it just offsets the first request each time, effectively rotating without taking any load indicators into account. Thanks! Gabriel Wicke
