tom lee wrote:
On Sat, Apr 26, 2008 at 9:20 AM, Roger Heflin <rogerheflin@xxxxxxxxx> wrote:
Because, if you keep writing to a corrupted filesystem you can end up
destroying the entire filesystem completely and lose *ALL* of your data and
that is worse.
I agree with you. That is why I think the OS should better off with
reboot "showdown -r -F now"
instead of mounting as read-only. if there is potencial disk problem,
you need to run this command anyway no matter what problems you may
find before rebooting.
If you automatically reboot, you make the problem go away, but don't have any
details on what happened (cannot log) so it *WILL* happen again, automatically
reboot helps recover from the issue, but results in the loss of all details that
could be used to fix the issue.
The problem is that it may or may not crash before it destroys the
filesystem competely, and if the OS is written robustly it should not crash
just because the filesystem tables are corrupted (and Linux has done some
testing with something that puts random data on the filesystem to make sure
that it does not crash in those random corrupted data cases).
From this perspective, I think microsoft way of crasing is a better
design. at least you know some wrong right away and reboot the
computer automatically can get it fixed.
That was not their design, MS tends to try to work around errors rather than
report the errors, so if then you get an error, it tries to cope and then
you get a completely unrelated cryptic error that really tells you nothing.
If if the crash said nothing useful to identify the failing component is it
useless, you have no idea what to fix, just crashing tells someone nothing.
If you had checked dmesg there should have been a clear error indicating
what happened, if all of the partitions on the filesystem were RO then I
would suspect that the disk itself quit talking, next time make sure to
check dmesg and see what it says.
ok. so it is too late to check since I already rebooted the OS?
Yeap, too late, if the root filesystem goes RO it does not leave any tracks
except in dmesg.
I have seen the RO remount a number of times on lots of different HW/kernel/dist
combinations, it is can be any number of issues, from a real hardware issue, a
hardware driver issue, a filesystem driver issue, a bios issue, a main kernel
issue, ... it has a lot of causes.
If it happens again, type "dmesg" and if possible save it someplace that it not
readonly (type sync a few times to make sure it gets saved-or put it on a flash
driver and if you have another device verify you have it), and then reboot.
Roger
--
fedora-list mailing list
fedora-list@xxxxxxxxxx
To unsubscribe: https://www.redhat.com/mailman/listinfo/fedora-list
