Re: Squid multithread

Yuri Voinov <yvoinov@xxxxxxxxx> · Mon, 14 Nov 2016 23:09:30 +0600

    14.11.2016 21:59, Eduardo Carneiro
      пишет:

      Yuri Voinov wrote

        http://wiki.squid-cache.org/Features/SmpScale

http://wiki.squid-cache.org/MultipleInstances

14.11.2016 20:22, Eduardo Carneiro пишет:

          Hi everyone!

I have a Squid 3.5.19 with dynamic content cache using url rewrite. It's
a
virtual machine (VMWare) with 2 quad-core processors each. Squid proccess
is
100% in one of my eight cores and the access it's getting very slow.

There exists any way to compile squid with multithread option? The idea
is
spread this processing beetwen the 8 cores.

Thanks.

--
View this message in context:
http://squid-web-proxy-cache.1019090.n4.nabble.com/Squid-multithread-tp4680488.html
Sent from the Squid - Users mailing list archive at Nabble.com.
_______________________________________________
squid-users mailing list

        squid-users@.squid-cache

          http://lists.squid-cache.org/listinfo/squid-users

        -- 
Cats - delicious. You just do not know how to cook them.

_______________________________________________
squid-users mailing list

        squid-users@.squid-cache

        http://lists.squid-cache.org/listinfo/squid-users

0x613DEC46.asc (2K)
&lt;http://squid-web-proxy-cache.1019090.n4.nabble.com/attachment/4680489/0/0x613DEC46.asc&gt;
signature.asc (484 bytes)
&lt;http://squid-web-proxy-cache.1019090.n4.nabble.com/attachment/4680489/1/signature.asc&gt;

      From what I understood I will need to use each instance in different ports,
right? This does not solve my problem. There are any way to separate one
instance for accesses and another for cache?

    What can workers share?

    Using Coordinator and
      common configuration files, Squid workers can receive identical
      configuration information and synchronize some of their
      activities. By default, Squid workers share the following: 

      Squid executable, 
      general configuration, 
      listening ports (but shared ICP and HTCP ports do not work
        well; see below), 
      logs, 
      memory object cache (in most environments), 
      disk object cache (with Rock Store only), 

        insecure cache
          manager statistics (detailed elsewhere), 

      SNMP statistics. 

    Cache indexes are shared
      without copying. Other shared information is usually small in
      terms of RAM use and is essentially copied to avoid locking and
      associated performance overheads. 
    Conditional configuration
      and worker-dependent macros can be used to limit sharing. For
      example, each worker can be given a dedicated http_port to listen
      on. 
    Currently, Squid workers do
      not share and do not synchronize other resources and services,
      including (but not limited to): 

      memory object cache (in some environments), 
      disk object cache (except for Rock Store), 
      DNS caches (ipcache and fqdncache), 
      helper processes and daemons, 

        stateful HTTP
          authentication (e.g., digest authentication; see bug 3517), 

      delay pools, 
      SSL session cache (there is an active project to allow session
        sharing among workers), 

        secure cache
          manager statistics (detailed elsewhere), 

        ICP/HTCP (works
          with a caveat: If multiple workers share the same ICP/HTCP
          port, an ICP/HTCP response may not go the worker that sent the
          request, causing timeouts at the requesting worker; use a
          dedicated ICP/HTCP port as a workaround). 

    Some SMP-unaware features
      continue to work in SMP mode (e.g., DNS responses are going to be
      cached by individual workers), but their performance suffers from
      the lack of synchronization and they require more resources due to
      duplication of information (e.g., each worker may independently
      resolve and cache the IP of the same domain name). Some
      SMP-unaware features break badly (e.g., ufs-based cache_dirs
      become corrupted) unless squid.conf conditionals are used to
      prevent such breakage. Some SMP-unaware features will appear to
      work but will do so incorrectly (e.g., delay pools will limit
      bandwidth on per-worker basis, without sharing traffic information
      among workers and without dividing bandwidth limits among
      workers). 

--
View this message in context: http://squid-web-proxy-cache.1019090.n4.nabble.com/Squid-multithread-tp4680488p4680490.html
Sent from the Squid - Users mailing list archive at Nabble.com.
_______________________________________________
squid-users mailing list
squid-users@xxxxxxxxxxxxxxxxxxxxx
http://lists.squid-cache.org/listinfo/squid-users

    -- 

      Cats - delicious. You just do not know how to cook them.

Attachment:
0x613DEC46.asc

Description: application/pgp-keys
Attachment:
signature.asc

Description: OpenPGP digital signature
_______________________________________________
squid-users mailing list
squid-users@xxxxxxxxxxxxxxxxxxxxx
http://lists.squid-cache.org/listinfo/squid-users