[squid-users] R: Re: problem reload configuration with workers

[Alex Rousskov]

On 09/27/2016 03:15 AM, amaury at tin.it wrote:
> I have resolve my problem changing the workers configuration from:
> 
> workers 4
> http_access allow localhost
> http_port localhost:
> 400${process_number}
> cache_dir aufs /var/spool/squid 16384 32 512
> 
> cache_dir rock /var/spool/squid/rock 16384 max-size=32768
> cache_dir aufs /var/spool/squid/squid-cache0${process_number} 16384 16 512 min-size=32769

The above combination of SMP-aware and SMP-unaware caches is unsupported
and probably will not work.


> to
> 
> workers 4
> http_access allow localhost
> http_port 
> localhost:400${process_number}
> cache_dir rock /var/spool/squid/rock 16384 max-size=32768
> 
> if ${process_number} = 1
>         cache_dir aufs /var/spool/squid/squid-cache01 16384 16 512 min-size=32769
> endif
> if 
> ${process_number} = 2
>         cache_dir aufs /var/spool/squid/squid-cache02 16384 16 512 min-size=32769
> endif
> if ${process_number} = 3 
>         cache_dir aufs /var/spool/squid/squid-cache03 16384 16 512 min-size=32769
> endif
> if ${process_number} = 4
>         cache_dir aufs /var/spool/squid/squid-cache04 16384 16 512 min-size=32769
> endif

The above is equivalent to

  workers 4
  http_access allow localhost
  http_port
  localhost:400${process_number}
  cache_dir rock /var/spool/squid/rock 16384 max-size=32768
  cache_dir aufs /var/spool/squid/squid-cache${process_number} 16384 ...

which is still an unsupported combination of SMP-aware and SMP-unaware
caches that probably will not work.

BTW, why do you give each worker a dedicated listening port (i.e.,
"400${process_number}")?

Alex.


> ----Messaggio originale----
> Da: rousskov at measurement-factory.
> com
> Data: 26-set-2016 16.12
> A: <squid-users at lists.squid-cache.org>
> Cc: 
> "amaury at tin.it"<amaury at tin.it>
> Ogg: Re: [squid-users] problem reload 
> configuration with workers
> 
> On 09/26/2016 08:02 AM, amaury at tin.it 
> wrote:
> 
>> I'm using squid 3.5.21-20160908-r14081 and for the first time 
> I'm 
>> using workers configuration. I have a problem:
>> when I reload 
> configuration (via init script)
>> suid -k reconfigure -f 
> /et/squid/squid.conf
> 
> I assume that by "suid" you meant "squid". If 
> yes, then the above
> command is a correct way to reconfigure Squid, 
> including SMP Squid.
> 
> 
>> the system kill squid-coord and squid-disk
> 
> 
> Does "the system" do more than run "squid -k reconfigure ..."?
> 
> * If 
> not, then "the system" does not kill squid-coord and squid-disk
> 
> (something else does).
> 
> * If yes, then you should fix your system 
> script. Perhaps it thinks that
> Squid died and tries to kill/restart it?
> 
> 
> 
>> So I have to remove pd 
>> file and lock files and restart squid.
>>
> Please, do it exist a solution 
>> to reload with restart squid?
> 
> 
> Reconfiguration should work "as is". If it does not work, file a bug
> 
> report with details such as your system command(s) and resulting
> cache.
> log and syslog output.
> 
> 
>> I tried also using something like:
>>
>>
> pid_filename /var/run/squid/squid-{proccess_number}.
> 
> Please do not do 
> that. SMP Squid is designed to work with a single
> configuration file 
> without SMP macros.
> 
> 
> Thank you,
> 
> Alex.
> 
> 
>