Re: Peers don't work ?

Miquel van Smoorenburg wrote:

> I found out why this is. At startup, all peers are dead. Then after
> some time all peers are _reported_ to be alive for a certain period.
> Then they are dead again.

Ah, you have a Squid with to low ICP query frequence ;-)

I have patches for this. See http://hem.passagen.se/hno/squid/, it
changes Squid to consider peers alive even if not queried since long.

> Indeed, I had an "always_direct allow cisnet" where "cisnet" is
> my local network .. but ofcourse you need to define that as
> a "dst" ACL, not a "src" ACL :/

Ah, there is the reason to the low frequency.
 
> The output from "peer cache statistics" could be improved though;
> if it had told me none of the peers were ever queried I would have
> found the solution a lot earlier.

It did, at least in the first printout you showed. There the number of
pings sent was 0, and why I asked you if it really worked in the
beginning. I have no clue why Squid would ever query peers siblings if
always_direct is always true, so I suspect you have some user or peer
where always_direct isn't true. It could also be something with digest
or netdb exchanges..

--
Henrik Nordstrom
Squid hacker