Usually, this means that (for whatever reason) the cache was unable to
connect to it's parent. Look in your cache_log. You should see a line
like 'TCP connection to ______ failed', at which point it will start
reporting the error, followed a little later by 'TCP connection to _____
succeeded'.
The failure to connect can be caused by (pick one):
* Network congestion, and dropped packets
* Momentary routing loss
* Shortage of available file-descriptors to the squid process on either
machine.
Once the connection with the parent fails, it is marked as 'down' for a
brief period, and not used. If you want to avoid this behaviour, and not
have squid mark the parent as 'down', but to keep trying to fetch from
it, edit src/neighbors.c
Find the subroutine peerCheckConnectStart()
Change the line that says: p->tcp_up = 0;
so that it says: p->tcp_up = 1;
DISCLAIMER: THIS IS NOT A CURE. THIS ONLY ALLEVIATES THE SYMPTOMS.
Examine the list of causes (above) carefully, and figure out which is
your problem (there may be others that I've missed, but I obviously
cannot think of any others at this point). Address that problem, and
you've fixed it. Ignore it, and at best this is just a band-aid patch.
It _does_ give you breathing space, to track down the problem, though,
which is AFAIK not with the squid code.
Richard Stagg wrote:
>
> Greetings,
>
> The system: Squid 1.1.17; Linux 2.0.32; 256Mb; 2Gb hard disc.
>
> The problem: Occasionally, the system will suddenly begin reporting
> ERR_CANNOT_FETCHes, with the recorded error that:
>
> protoCantFetchObject: FD xx No peers to query and the host is beyond your
> firewall.
>
> [I use a different box to check the firewall's still alive - it always
> is.]
>
> This continues for a couple of minutes, and then reverts to its normal,
> functioning, state. This is a recent bug; the cache in question has run
> for months without any problems, and nothing on the box has changed (to my
> knowledge) during that time.
>
> Any suggestions greatly appreciated.
>
> Richard Stagg
>
> -----------------------------------------------------------------------------
> ** Disclaimer: My opinions have nothing to do with CSC. Keep it that way. **
> Richard Stagg - Internet Analyst and General Duties Unix Hacker
> 0118 933 1491 / 01252 383153 / 0378 300720 / stagg@lentil.org / rjs@bae.co.uk
> -----------------------------------------------------------------------------
-- -----BEGIN GEEK CODE BLOCK----- Version: 3.1 GAT d- s++: a C++++$ UL++++B+++S+++C++H++U++V+++$ P+++$ L+++ E- W+++(--)$ N++ w++$>--- t+ 5++ X+() R+ tv b++++ DI+++ e- h-@ ------END GEEK CODE BLOCK------Received on Mon May 18 1998 - 06:21:59 MDT
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:40:13 MST