On 11/22/2010 11:49 PM, Amos Jeffries wrote:
On 17/10/10 19:40, Arthur Titeica wrote:
On Wednesday 04 August 2010 16:55:46 Amos Jeffries wrote:
Zeller, Jan (ID) wrote:
It looks like that message only occurs on a reconfigure. Does "-k
restart"
after the config change work?
Amos
hmm the change applies once squid is restarted but now I am getting :
010/08/04 08:21:20 kid3| commBind: Cannot bind socket FD 12 to [::]:
(13)
Permission denied .
.
squid is running as
cache_effective_user proxy
cache_effective_group proxy
squid processes are running but no listening port. Any clue why this
happens ?
Nothing I know about should lead to a kidN using bind on [::] or
0.0.0.0.
Maybe Alex has a clue.
cc'ing to squid-dev where beta release problems really need to be
discussed. Please followup there.
Amos
Sorry for bumping this thread but I see the same with 3.2.0.2-20101016.
Something is trying to bind to [::] and the proxy is refuzing conections.
The thread title suggests that the 'workers' setting is ignored but I
guess it's not as I see the 4 squid processes with ps.
Since that earlier response I've learned a thing or two about how the
kidN work. With "workers 2" the "kid3" is the coordinator process which
opens shared ports and passes them to the other kids for sharing.
The lack of port number in the address being opened and bound seems to
be the problem.
Do you have all http_port configured inside if..endif for specific
workers, or some shared ones?
Are you running Squid as root or a less-privileged user?
Is IPv6 disabled on the box?
I am probably missing a lot of background here (I did not see this on
squid-dev), but I would recommend:
a) Checking that everything works with one worker (default) or when
Squid is started in no-daemon mode (squid -N ...).
b) Posting your squid.conf and cache.log after setting debug_options to
ALL,9 and reproducing the problem. It may be better to post this to
bugzilla.
Thank you,
Alex.
