Gergely Madarasz wrote:
> So what I'd need here is: discover why ICP fails on both hosts at the same
> time (one is two hops away, the other three, and while using them
> round-robin-ed, there is no problem at all), fix this, and then configure
> ICP and have my cache use both parents when available for load sharing...
Or you could use lower connection timeouts for your peers to speed up
the dead peer detection on link failures. A few seconds timeout should
be sufficient for any decent peer connection. Squid-2.2 patch
implementing this is available from my Squid page, or wait for
Squid-2.3.
You shouldn't need to use ICP it it is only for monitoring if the peer
is alive. TCP connections does that job just fine.
-- Henrik Nordstrom Squid hacker http://hem.passagen.se/hno/squid/Received on Wed Jan 05 2000 - 18:29:51 MST
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:50:14 MST