Sorry for wasting everyone's time a bit...but I did want to try to
focus in on what is really a problem I can't figure out
and what isn't before filing a bug in Bugzilla.
Thank you! I nailed it...at least in terms of Unix Domain Sockets.
Squid is up and running with 3 workers and not sharing memory.
At this point I'll ask if anyone has a FreeBSD box handy?
If so, could you please run "sysctl -A" and let me know what the
following are set to.
Currently I'm using the settings from the squid Wiki and have little
idea what to set them to on Mac OS X.
net.local.dgram.recvspace: 262144
net.local.dgram.maxdgram: 16384
I'm going to give what I have a little time to "settle in" and then
get back to someone else who wanted a working SMP configuration also.
Regards,
On Sat, Mar 9, 2013 at 12:07 PM, Alex Rousskov
<rousskov_at_measurement-factory.com> wrote:
> On 03/09/2013 09:16 AM, Adam W. Dace wrote:
>> This is Mac OS X v10.8.2.
>>
>> Yeah, I'm having all sorts of problems with shared memory.
>
> and with inter-process communication (IPC) via Unix Domain Sockets (UDS):
>
>> With "memory_cache_shared off", I get a different error where the
>> child processes are unable to register with the coordinator process.
>
>
>
>> Maybe there's some bizarre Mac-specific command I have to issue to
>> enable shared memory support?
>
> It is likely that you have that support enabled but Squid needs to do
> something special and/or you need to tune settings such as UDS message
> sizes and permissions for everything to work smoothly. Every platform
> has its own shared memory and IPC oddities, unfortunately. Some known
> ones are documented at
> http://wiki.squid-cache.org/Features/SmpScale#Troubleshooting
>
>
>> Any help would be appreciated.
>
> You need to figure out what system call is failing first and go from
> there. I recommend disabling shared memory and making sure IPC is
> working first -- you will need IPC anyway.
>
> Some of the things you can try are:
>
> * Running Squid with debug_options ALL,9 and figuring out why kid
> registration does not work. Often, the OS limit for UDS messages is too
> low. There should be some error messages in ALL,9 cache.log about that,
> but they are difficult to find and interpret.
>
> * Running strace/ktrace (or Mac equivalent) and analyzing error codes of
> the failing system call (by matching ALL,9 cache.log with the trace).
>
> None of the above is easy or straightforward, unfortunately. And fewer
> folks are going to volunteer their help since you are running on a
> non-free OS (I am not saying this in a negative way -- just stating the
> facts). You may want to file a bug report to post your logs, ask for
> help, and keep a better record of your progress.
>
>
> Good luck,
>
> Alex.
>
-- ____________________________________________________________ Adam W. Dace <colonelforbin74_at_gmail.com> Phone: (815) 355-5848 Instant Messenger: AIM & Yahoo! IM - colonelforbin74 | ICQ - #39374451 Microsoft Messenger - colonelforbin74_at_live.com Google Profile: http://www.google.com/profiles/ColonelForbin74Received on Sat Mar 09 2013 - 18:48:39 MST
This archive was generated by hypermail 2.2.0 : Tue Mar 12 2013 - 12:00:04 MDT