regular fsck runs are too disturbing

[HggdH]

On Sun, 2007-12-02 at 22:55 +0000, (=?utf-8?q?=60=60-=5F-=C2=B4=C2=B4?=)
-- Fernando wrote:

> Dane , you can manually bypass this by using tune2fs, and disable the fsck on your server.

Yes, indeed this will do the trick. But it requires knowledge of some
quite arcane utilities -- not usually what the casual user has --, and
bypasses the basic issues:

1. fsck takes an inordinate long time for large filesystems;

We distribute Ubuntu with the installation by default in one single
monolithic filesystem (and most other distributions will do the same).
Of old this was no biggie, since the disks were (relatively) small. But,
nowadays, we usually get harddrives in excess of 100G. 

Very few of us (based on my experience) will partition the HD. I have
had issues on Ubuntu on this (I *do* run many partitions), with software
updates putting critical system utilities in /usr/[s]bin instead
of /[s]bin -- which causes some rather bad errors on boot (/usr is a
mount point on my systems)

2. a generic ~30 mounts per check is too short an interval.

Although this is probably good enough for desktop systems, it breaks
fast on laptops. I, for example, boot my laptop at least twice a day --
so, on my personal case, I will have a forced check in (usually) less
than 2 weeks time. If I were to be running a single fs, it would take
about 25 minutes for it to complete. Fortunately for me, since I broke
my install in many filesystems, not all of them get done at the same
time. 

[as an example, I have seem my wife get out of her laptop in disgust
when such a check started. And, of course, blast me for that :-)]

3. taking out the check is potentially dangerous in the long run.

A direct question here is: how long can such a check be postponed? This
question has not yet been answered, and we have people either disabling
(via tune2fs or friends), or putting in some arbritary values.

What we need is some consensus on how to deal with it.

-x-x-x-x-x-x-

I am guessing what we would need here is a reanalysis of how the checks
are done, and what could be changed to minimise the impact of such
checks. I would expect changes in the filesystems also.

Perhaps a way would be a routine to prompt the user for a check next
reboot, and be increasingly more vocal if the user keeps on postponing
the check: 

* This system has run for xxx (days|months|boots|whatever)
* without a FS check. Do you want this check performed
* next boot? 
*
*  [ ] Yes [ ] No [ ] postpone for now

And then the routine would set a flag to be read by something next boot.

 

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
URL: <https://lists.ubuntu.com/archives/ubuntu-devel-discuss/attachments/20071204/b1c22039/attachment.sig>