Dear Alpha Guru's,
I've been happily running Squid-1.1.20 for the past few months with a
small (256Mb) AdvFS fileset on a 4Gb disc. There's only a few client so
the size is quite adequate. This is on a Alpha Workstation 500au running
DU4.0D. Today I installed a new Seagate 4Gb drive, created a new 4Gb
domain, created a new soft-limited 600Mb fileset for the cache. Stopped
squid, vdump/vrestore'd the cache files (approx. 25,000 files in 240Mb)
from the old fileset to the new. Umount'd old fileset, mounted up the new
fileset. Started up squid. Seemed to run fine for the past 10 hours. I
check it now and I'm getting these errors -
kern.log:
vmunix: /opt/cache: write failed, file system is full
last message repeated 61 times
> df /opt/cache
Filesystem 1024-blocks Used Available Capacity Mounted on
domain2#cache 600000 68921 531079 11% /opt/cache
Squid is complaining about "disk write error: (28) No space left on
device", and purging objects. As you can see above I've only got 70Mb left
in the cache!
I'm about to give up on AdvFS and partition the 4Gb into two slices, one
3.5Gb for 'normal' AdvFS use, and 500Mb UFS for squid cache. Is this the
best solution? Am I just not tuning AdvFS it right? Or is AdvFS just bad
news for this type of disc/file activity?
I've had no problems on another machine running DU4.0B and a 1Gb AdvFS
Squid disc cache (that machine also has a 1Mb prestoserve cache). But I
have had similar problems on a DU4.0B and a 8Gb AdvFS (2x4Gb discs) cache
server - I've managed to keep that machine working until I get the chance
to rebuild with UFS.
_____________________________________________________________________________
Andrew "Alf" Leahy, phone: 02 47 360622
Unix Systems Administrator, mailto:alf_at_cit.nepean.uws.edu.au
School of Computing & IT, UWS Nepean
http://www.cit.nepean.uws.edu.au/~alf/
Received on Tue Jun 02 1998 - 13:21:14 NZST