Monitoring Mounted EXT3 Filesystem Consistency

[Allan Morrison]

I have recently joined a group responsible for a 7X24 environment for
about 100 different client environments. We use EXT3 running under
Oracle Unbreakable Unix. My question is about insuring filesystem
consistency. In my previous lives the filesystems would come unmounted
for various maintenance windows often enough to insure that an fsck
would be run several times per year. This was good.

In this environment there are the dreaded 'Service Level Agreements' and
administrative overhead for unscheduled maint windows. We basically
don't get to take the environment offline without a compelling and
imminent disaster.

So my question is: "Does anyone have any methods or strategies for
overseeing the consistency of a filesystem (ext3) that is not permitted
offline without documented proof of a problem that needs a addressing?"

So far the idea that I have is to run and e2fsck -n on them weekly and
monitor the results of the first three passes seem pretty believable.
The Inode and block counts seem to come out wrong pretty much every
time, even on a freshly minted unused 1gb filesystem I made for testing
this idea. I even tried syncing the filesystem first, and that didn't
even change the #'s the wrong counts found.

This might be good enough, it seems like an improvement to me.

Comments, questions, suggestions?

Allan Morrison