Re: [squid-users] SQUID2.5.D/lx2.4.4 - Peer failover - how to config squid to fail over to direct queries if upstream is offline? (NON-ICP PEER)

Ben Ryan wrote:
>
> Thanks for your help Robert, and the quick reply.
>
> ==
> A few further questions for the list, anyone
> who can reel off some info on the below queries:-
>
> Are these all the timeouts relevant to a non-ICP upstream peer?
> #squid.conf
> connect_timeout (for direct http gets?)

yes.

> peer_connect_timeout (when this times out, does squid go direct?)

yes.

> read_timeout

yes.

> >From the docos it's not clear which parameters apply to which
> situation; is connect_timeout a tcp timeout relevant to
> squid->webserver connections or user->squid connections?
> Should I check this with the dev's or duane via the faq?

When Squid makes a connection somewhere. i.e. to origin servers or peer
caches.

The peer_connect_timeout refines this for peers, and the cache_peer
connect-timeout option refines it for individual peers.
 
> One piece of important info I couldn't discern;
> How would one set the post-parent-failure retry interval?

There is none for TCP-only peers. The probing piggy-backs on your normal
requests. If peer selection finds dead TCP-peers then a TCP probe is
started, attempting to connect to the peers (at most one such probe at a
time is running). When a connect is successfull the peer will be
considered alive again.

--
Henrik Nordstrom
Squid Hacker