On Tue, 27 Apr 2004 Hanno.Wagner@boerse-stuttgart.de wrote:

> Hi,
>
> I have here a Netra-machine with Solaris 8 and 1GB RAM. This machine is
> serving as mailhost/mailscan-machine and Webproxy. Today,squid stopped
> because it claimed not to have enough storage space to store and it tried
> to rebuild the store database.


Make sure it is not running into a 2GB file size limit on any of the log
files..

Also on Solaris situation can be worse if the filesytem is not tuned to
optimize for space. The problem is that df includes block fragments in the
free space report, but for writes to larger files (>3KB or so) to succeed
there needs to be full blocks available, not only fragments. A cache that
has been running for some time with high usage of the filesystem with the
filesystem tuned for "optimize for speed" can cause quite servere free
block fragmentation as web traffic typically coinsists of very many small
files.. To detect if this is the case you can use dd

dd if=/dev/zero of=somefile

if this stops early while there still is lots of free space reported by df
and ulimit -a does not say filesize is limited then your filesystem free
space is most likely too fragmented to be useable. It helps reducing the
cache size, but you may need to first delete some content to get Squid
running again.


Regards
Henrik