From: Rudolf Gabler (rug@usm.uni-muenchen.de)
Date: Fri Apr 07 2006 - 02:27:28 EDT
Hi all,
we run TruCluster latest OS (patchkit) and observed the following:
. propably due to an increase storage (number of files) the var-filesystem
started to overwrite itself, i.e. we found system-logs in the
var/spool/mail/someuser files (/var/spool/mail is extra mounted by a
cluster_var#mail fileset, i.e. the overwriting occurs also beyond
filespace-limits)
. the cluster_var fd was already created with a mkfdmn -x 8192 -p 230400
option
. the filespace was far below the limit (34%)
. we stopped immediately the cluster and run fixfdmn cluster_var in single
user mode.
. fixfdmn complained 10 times that it had no free pages i.e. run out of
inode space but finished "properly", i.e it managed to repair the var
filesystem. O.k. to tell the truth it removed everything, no directories or
files on the var filesystem left, full space now!
. There was no indication that the filesystem would run out of any limits,
no error message in any log. It would be friendly, if the admin is telled:
hey your inode limit is below 10 % or something similar.
. A filesystem utility which removes everything should be declared: "only
for hazards"
Unfortunately we lost all the user data in the mail-spool (o.k. we don't
save this area because of the rapid changing kind of files and warn any 3
weeks the users, that this area isn't save).
Any similar experiences?
Rudolf Gabler
This archive was generated by hypermail 2.1.7 : Sat Apr 12 2008 - 10:50:29 EDT