ags@cernvax.BITNET (10/31/86)
Dear Sirs, I would be most obliged if you could publish the following request
for information in your modcomputers.vax series.
Alan Silverman
CERN, Geneva
Disc Corruption Problem
-----------------------
Request for information:
Some 5 times since June, we have experienced a serious disc corruption problem
on our clustered 8600. First one or two files have "unsupported file structure"
and this gradually spreads to more and more files on the disc. It turns out
that some file extents are not properly marked as being used and/or parts of
the bit map have been zeroed! The problem is very contagious. It is not always
the same disc which gets clobbered.
We also have a stand-alone single-disc 780 which has seen the same problem
once or twice. Once per week, we stop this 780 and mount its disc on our cluste
HSC for backup.
We see two possible causes -
a) hardware - the corrupted discs share the same HSC.
b) VMS. We are running 4.3. DEC applied a patch to RMS but we got hit again.
Now they claim that the ACP cache parameters in SYSGEN should not be altered
from the defaults. We did not change them, AUTOGEN did.
The first cases happened shortly after system crashes and we suspected a
lost cache problem. The RMS patch was supposed to fix a problem where a
process gets aborted by a crash or a STOP/ID. However, the last occurrence
was after the installation of the patch and seemingly unrelated to a crash.
Has anyone else seen or heard of such an error? Anyone else out there with a
large heavily loaded cluster system and non-default ACP parameters?
Alan Silverman, CERN, Geneva
Address - DD Div, CERN, 1211 Geneva 23, Switzerland
Mail to alan@vxcrna.cern (EAN)
vxcrna::alan (HEP DECNET)
...!mcvax!cernvax!ags (UUCP)
alan@cern.vxcrna (CBS)