Re: workers binding to [::]

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.
Received on Thu Nov 25 2010 - 22:14:17 MST