[Lias] squid problems

[Jonathon Coombes]

On Thu, 2002-12-19 at 18:04, ken_yap at users.sourceforge.net wrote:
> >BTW, if you're running Red Hat's standard squid package, the log files should
> >get rotated automatically by logrotate.  If they're not, something's wrong with
> >your logrotate config.
> 
> While a disk full proved to be the case, what I didn't mention in my
> little story is that the virus managed to push store.log up to 2GB in
> the space of 24 hours. The point of my story is not whether logrotate is
> running or not, that's boring, but that sysadmins should check that
> logrotate is running *often enough*.  The default for squid logs is 1
> week IIRC but on high volume servers you may need to reduce this
> interval. store.log reaching 2GB is something you will see noted in
> cache.log, but cache.log reaching 2GB is not something you will see
> logged, obviously. The only clue you'll see is the magical size of
> 2147483647 for the log file. I'm sure a few of the mysterious "squid
> won't restart, so I reinstalled and the problem went away" are due to
> this situation.

In situations where the traffic on squid is very high and/or may be
harder to determine over a set interval, you could change the manner
of logrotate to a fixed size. That is, it does not need to be set to
a time period for rotation, but could be set to rotate once it reaches
100MB for example.

This is also very  useful if you are debugging squid and the logs are
filling up extremely fast :-)

Regards,
Jonathon