[ale] Fwd: periodic fsck was Re: [patch] ext2/3: document conditions when reliable operation is possible

[Greg Freemyer]

On Tue, Nov 10, 2009 at 8:12 AM, Ed Cashin <ecashin at noserose.net> wrote:
> On Tue, Nov 10, 2009 at 7:55 AM, Rev. Johnny Healey <rev.null at gmail.com> wrote:
>> I believe the reason that you can't fsck a mounted fs is because of the
>> chance of writes during the fsck.  But when you make a snapshot, it doesn't
>> get mounted until you explicitely mount it somewhere.
>
> Yes, it should be just like running fsck on an unmounted filesystem after
> a power failure (no clean shutdown), if I understand correctly.

No.  Part of the lvm snapshot process is to tell the filesystem clean
up the journal etc.  It should be fully applied and empty at the time
of the snapshot.

> In that situation, you can get some errors reported for an otherwise healthy
> filesystem, just because it wasn't left in a 100% consistent state.
> If the power
> hadn't gone out, those inconsistencies would have been corrected in the
> following seconds.  With a snapshot, though, those same inconsistencies
> will appear, even though there's no persistent problem with the filesystem.
>
> So I would think that unless the fs was unmounted before the snapshot
> was taken, the
> admin wouldn't be able to tell between errors related to the fact that the
> snapshot was taken on a live filesystem and errors related to serious and
> persistent problems.

Ted Tso would know far more that you or I, so if he recommends it I
suspect it is a good plan.

The thread is at:

http://markmail.org/thread/2tu3brz74zuafxqm

The message at:

http://markmail.org/message/dn74kxsb7wigu643

Greg