On 10/01/2014 11:24 PM, Amos Jeffries wrote: > We already have client_lifetime directive which is documented as > doing exactly what this new directive does. > (<http://www.squid-cache.org/Doc/config/client_lifetime/>).
No, client_lifetime is not documented (and is not meant) to do what we want pconn_lifetime to do. Client_lifetime is supposed to kill client connections exceeding the configured lifetime, _regardless_ of the client (and client connection) state. We do not want that for pconn_lifetime which only affects _idle_ connections. Moreover, pconn_lifetime is meant to apply to all connections, not just the connections from a client. > * change your new directive "pconn_lifetime" to be > "client_pconn_lifetime" since it is specifically scoped to client > browser connections. pconn_lifetime is not exclusive to client browser connections. Ideally, it should apply to _all_ Squid connections that are in an "idle persistent connection" state. We are not there yet, but we already support origin connections AFAIK. > * update tunnel.cc client connection timeouts to use > conn->timeLeft(). - maybe with some new tunnel-specific lifetime > config item. But I think your new config lifetime is appropriate > for now at least. There is no concept of "idle persistent connection" for TCP tunnels so we should not apply pconn_lifetime to them right now. Hope this clarifies, Alex. _______________________________________________ squid-dev mailing list squid-dev@lists.squid-cache.org http://lists.squid-cache.org/listinfo/squid-dev