Hi Amos,
Thanks a lot for your help!
Unfortunately I had to switch back to aufs stores instead of rock stores, as I encountered another problem. I started to get:
Worker I/O push queue overflow:
communication with disker may be too slow or disrupted
Therefore, I added "max-swap-rate" and "swap-timeout" to my rock cach_dir's as follows:
cache_dir rock /mnt/cachedrive2/verysmall 100000 max-size=32000 max-swap-rate=200 swap-timeout=300
cache_dir rock /mnt/cachedrive3/verysmall 100000 max-size=32000 max-swap-rate=200 swap-timeout=300
cache_dir rock /mnt/cachedrive5/verysmall 100000 max-size=32000 max-swap-rate=200 swap-timeout=300
But I ended up not being able to start squid at all. There were again the same log entries complaining about shared memory:
FATAL: Ipc::Mem::Segment::open failed to shm_open(/squid-squid-page-pool.shm): (2) No such file or directory
Squid Cache (Version 3.3.8-20130816-r12602): Terminated abnormally.
Since this is a production site, I had to switch back to aufs!
Best regards,
Firas
----- Original Message -----
From: Amos Jeffries <squid3_at_treenet.co.nz>
To: squid-users_at_squid-cache.org
Cc:
Sent: Sunday, August 18, 2013 10:55 AM
Subject: Re: [squid-users] Two Potential Bugs When Using Rock Stores!
On 18/08/2013 10:25 a.m., Golden Shadow wrote:
> Hi!
>
> If I configure more than 3 rock stores, I get the following cache.log entry and squid is stopped abnormally:
>
> Squid Cache (Version 3.3.8-20130816-r12602): Terminated abnormally.
>
> FATAL: Ipc::Mem::Segment::open failed to shm_open(/squid-squid-page-pool.shm): (2) No such file or directory
> Squid Cache (Version 3.3.8-20130816-r12602): Terminated abnormally.
>
>
>
> Another bug is that WCCP does not work when using "wccp2_rebuild_wait on", while having rock stores. I think squid is not sending "Here I AM" message to the WCCP router as the router does not show any WCCP clients!
>
> My questions are:
> 1- Is it important to use "wccp2_rebuild_wait on"? Will there be any bad results, if I set this directive to off?
Maybe yes, maybe no. The side effect is that your proxy will operate as
a proxy without a cache until the cache index is finished loading. That
time varies for different cache types and sizes and whether the proxy
got shutdown properly earlier. If the amount of upstream traffic your
proxy will use for that time is more than your upstream link can handle
your users will see degraded service.
> 2- I'm only using one squid worker, so will using rock stores to cache files less than 32 KB still give any performance improvements over using aufs stores for such files?
Yes rock offers speed benefits over AUFS by collating small objects into
larger blocks and removing disk I/O overheads on objects which are
requested in groups. For example; all objects on a page grouped together
and HIT on first one loads them all into memory for quick followup HITs).
Amos
Received on Mon Aug 19 2013 - 09:01:12 MDT
This archive was generated by hypermail 2.2.0 : Mon Aug 19 2013 - 12:00:26 MDT