[squid-users] DiskThreadsDiskFile::openDone squid 3.5.0.4

[Tory M Blue]

On Thu, Feb 2, 2017 at 3:51 PM, Amos Jeffries <squid3 at treenet.co.nz> wrote:

> On 3/02/2017 7:56 a.m., tmblue at gmail.com wrote:
> > asnani_satish wrote
> >> This happens when size specified in cache_mem >= cache_dir
> >> Example:
> >> cache_dir aufs /var/spool/squid 1000 32 512
> >>    implies 1000 MB physical disk space allotted for cache in specified
> >> directory
> >> cache_mem 900 MB
> >>    cache size to be used by squid which must be less than the size
> >> specified in cache_dir directive.
> >>
> >> Dont forget to restart squid
> >>
> >> We cannot just ignore the error because if this error keeps on occurring
> >> the performance of squid degrades.
> >
> > Are you sure???
> >
> > i'm getting this error when turning back on a squid system that I had
> down
> > (out of production for 2 days). My config does not indicate an issue that
> > you advise to be the cause of the error.
> >
> > cache_mem 1 GB
> > cache_dir aufs /cache 65000 16 256
> >
> > I'm not seeing any issue but the error is rampant at this point
> >
> > 2017/02/02 10:58:59 kid1|     /cache/07/49/003749FE
> > 2017/02/02 10:59:00 kid1| DiskThreadsDiskFile::openDone: (2) No such
> file or
> > directory
> > 2017/02/02 10:59:00 kid1|     /cache/05/3E/00053EA0
> > 2017/02/02 10:59:00 kid1| DiskThreadsDiskFile::openDone: (2) No such
> file or
> > directory
> > 2017/02/02 10:59:00 kid1|     /cache/0D/EB/005DEB19
> > 2017/02/02 10:59:00 kid1| DiskThreadsDiskFile::openDone: (2) No such
> file or
> > directory
> > 2017/02/02 10:59:00 kid1|     /cache/02/27/0002273A
> >
>
> Squid uses a journal file swap.state to store changes to the cache. On
> restart it is used to reconstruct the cache index fast. If you cause
> Squid to shutdown very fast (kill -9, systemd crashing it, or
> shutdown_lifetime too short) then it does not have time to fully update
> the journal entries.
>
> When it gets restarted after one of those abrupt shutdowns one of these
> messages (not errors!) gets logged for each entry which the journal
> indicates are present - but are not really there because their removal
> was not recorded on shutdown.
>
> NP: There are also things in the cache being orphaned then just
> overwritten because the journal does not record them as existing. But no
> messages about that because its not detected.
>
> These messages should decrease exponentially proportional to your
> traffics normal HIT rate. But in a very large cache like yours it still
> may take many hours to become noticably slower and days to disappear
> entirely. That is of course assuming your Squid does not have another
> restart event to begin the process all over again.
>
> Amos
>

As always thanks Amos

These are systemctl stop squid and systemctl start squid, so nothing
dramatic or nasty, should be shutting down cleanly

Tory
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.squid-cache.org/pipermail/squid-users/attachments/20170202/d29bf8c7/attachment-0001.html>