dean.scothern@wwgsolutions.com wrote:
> I'm using squid-2.2STABLE5 across a congested WAN. Our provider has set the
> routers so that icmp packets are the first to be discarded. As a result I get
> regular dead_peer detection when its not.
Don't use ICP or ICMP pinging on a congested pipe.
> Also I'm a little confused about connect-timeout in the cache_peer line, what
> are the units? Is it similar to dead_peer _timeout?
The connect-timeout cache_peer option is how long Squid will wait for a
TCP connection to that peer. It is a per peer-specific version of
peer_connect_timeout.
dead_peer_timeout is used in ICP pinging, not exactly related to
connect-timeout.
Squid has two methods for declaring a peer dead:
a) ICP is used, and no ICP replies have been received within
dead_peer_timeout.
b) 10 TCP connections to the peer has failed in a row.
A peer is considered alive if both of the above is false.
-- Henrik Nordstrom Squid hackerReceived on Tue Feb 08 2000 - 17:36:13 MST
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:51:01 MST