http://brainstorm.ubuntu.com/item/11/
Possible solutions and further explanation:
http://micrux.net/?p=52

[1456 votes] Solution #1: Auto-generated solution of idea #11


]]> en-us Thu, 28 Feb 2008 14:28:23 +0000 Fri, 23 Oct 2009 06:51:06 +0000 QAPoll module http://brainstorm.ubuntu.com/idea/11/ Thu, 28 Feb 2008 14:36:41 +0000 Thu, 28 Feb 2008 14:50:55 +0000
https://blueprints.launchpad.net/ubuntu/+spec/partition-management
https://wiki.ubuntu.com/DesktopTeam/Specs/PartitionManagement

Priority: Low
Definition: Approved (Needs guidance)
Implementation: Started
Series goal: Accepted for hardy]]> Thu, 28 Feb 2008 15:05:03 +0000 Thu, 28 Feb 2008 15:53:07 +0000 OS is installed on physically separate hard drive.]]> Thu, 28 Feb 2008 16:37:33 +0000 Thu, 28 Feb 2008 17:24:57 +0000 Thu, 28 Feb 2008 22:17:00 +0000 Thu, 28 Feb 2008 23:49:17 +0000
I think running fsck periodically is important, but being able to postpone the check until next boot is also a good thing.]]> Fri, 29 Feb 2008 00:02:36 +0000 Well, the only thing that's lame is the fact I had to spent a couple of hours researching this topic to decide how to deal with it. It does not help the impression that in Linux everything requires time to figure out.

My suggestions is why not put a dialog screen after logging in that says "Hey you've booted more than 30 times? This is what the risk is. This is what we suggest you should do. And this is how you can customize how often this message should appear.

]]> Fri, 29 Feb 2008 00:03:45 +0000 Fri, 29 Feb 2008 00:36:05 +0000
I always run the system off a small partition for this reason (to minimise scan time) but that does lead to other problems.]]> Fri, 29 Feb 2008 00:37:14 +0000
Sorry you had such trouble. But it is lame that it took that long to find a solution. I don't think it should have taken more than 1 minute of your time to do a google search on what file to edit. Hell, including the time to edit the file and the google search should take you less than 1 minute.

Now if you're talking about "how to decide", well that's what lots of research is for. You can't really come to a good decision with only a little bit of research.

However, I do agree that it will be nice in Hardy that this is able to be bypassed. So, since this is really already fixed in the next release... can we remove this brainstorming option and get some better ones up here?]]> Fri, 29 Feb 2008 01:30:44 +0000
I can't think of a good reason that the filesystem integrity can't be checked while it is mounted, except that fsck does not implement this.

It would be even better if the filesystem could be _repaired_ while mounted, but one step at a time..


]]> Fri, 29 Feb 2008 01:32:34 +0000 Fri, 29 Feb 2008 01:40:24 +0000
How about an option to mount the disk as read-only if we skip fsck. JUST to be on the safe side - and you can still continue with your presentation.]]> Fri, 29 Feb 2008 01:57:36 +0000 ]]> Fri, 29 Feb 2008 02:11:12 +0000
You can always use another filesystem which does not need to be fsck'd that much (is it "that much"?) like reiserfs or jfs.

I consider jfs a fast and reliable filesystem in ubuntu linux, try it.]]> Fri, 29 Feb 2008 02:31:07 +0000 Fri, 29 Feb 2008 02:59:32 +0000
I agree that the best implementation would be to check and fix a filesystem on a running system in the background.
]]> Fri, 29 Feb 2008 03:23:44 +0000 Fri, 29 Feb 2008 03:27:12 +0000 Fri, 29 Feb 2008 04:13:33 +0000
Having a machine not ready for use only implies you did not allot enough time to prepare for your presentation or task.

I'm surprised no one has commented about using tune2fs to increase or decrease the mount count.

]]> Fri, 29 Feb 2008 04:42:49 +0000 Fri, 29 Feb 2008 04:57:21 +0000 Fri, 29 Feb 2008 05:02:01 +0000 Fri, 29 Feb 2008 05:38:20 +0000 Fri, 29 Feb 2008 07:38:35 +0000 Fri, 29 Feb 2008 08:11:29 +0000
Anyone seriously considering this need to read the following:

http://kerneltrap.org/Linux/Data_Errors_During_Drive_Communication

(I think that's the one I was thinking of, I didn't read the whole thing.)

Ya' know, breaking all the rules and incorporating ZFS into the linux kernel would be par for the course for Ubuntu at this point . . .

]]> Fri, 29 Feb 2008 08:16:05 +0000 Fri, 29 Feb 2008 10:09:46 +0000 Fri, 29 Feb 2008 11:24:27 +0000
I'm glad it can be aborted with [Esc] in Heron, although for me- it is important that it run.]]> Fri, 29 Feb 2008 13:49:28 +0000 no, the fsck must be done with the partition mounted read only, something that defeats completelly the "do it while booting" idea: read only system means "where's my /tmp" and problems like that for a lot of programs.

Maybe this is too much of a hack, but... What about making the schreduled "check of the 30 boots" on shutdown, instead of on boot time?

I usually are in a bigger hurry when starting the laptop, and once the hibernate/sleep problem is resolved it whould make more sense.]]> Fri, 29 Feb 2008 15:26:52 +0000
if boot_count > 25
print "disk has been mounted $boot_count times. Run fsck now? Y/N"
if Y
run fsck
if N
exit

Now how hard is that and where do I put the fix?]]> Fri, 29 Feb 2008 19:15:05 +0000 Fri, 29 Feb 2008 21:50:03 +0000 Fri, 29 Feb 2008 21:58:22 +0000 Sat, 01 Mar 2008 00:08:25 +0000 Sat, 01 Mar 2008 00:15:18 +0000
I cannot see why fsck can be done in parallel with the rest of the boot.

Of course, in reality it probably will mean decreased performance. However I think it is possible. Where's /tmp? Just use tmpfs or a ramdisk. It's how all the LiveCD distros boot. Problem solved.]]> Sat, 01 Mar 2008 03:28:18 +0000
On the 29th boot:
the system boots normally. except when after X loads, an applet pops up saying some thing like "A normal/regular disk maintenance check will be preformed on your next boot/reboot" the applet provides an option too:
1) chance this to to run on your next shutdown
2) to wait N days/reboots before running the maintenance
3) to stop running this maintenance forever(Giving a big warning of this bad choice...)
4)....

on boot 30:
if option (1) was selected fsck is run.

on boot 29+N:
if option (2) was selected fsck is run.
*********************************************
Note:::
when fsck is running also provide ALL the options listed above(1,2,3,4...) and offer the option to run fsck on next boot(crt-C type interruption)..

]]> Sat, 01 Mar 2008 08:00:52 +0000
I second the suggestion of a shutdown fsck.]]> Sat, 01 Mar 2008 09:33:18 +0000 Sat, 01 Mar 2008 20:26:51 +0000 Sat, 01 Mar 2008 20:46:53 +0000 Sat, 01 Mar 2008 21:02:31 +0000
-Canceling the check is something that will be in Hardy 8.10 AFAIK. (I have seen it in the alpha version and while it didn't work properly I expect that to be fixed in the final version. Should be a non-issue)

-The fact that the check is made every 30times might be a problem. Some ppl hardly never shutdown (hibernation/suspend).
Some do it everyday (suspend doesn't work...).

How about having the check interval in days (I remember reading this can already be easily done so what should be the default is the issue).

I don't know if it could be (easily) possible to use the computer uptime a way to determine when the check need to be done. (?)]]> Sun, 02 Mar 2008 13:20:22 +0000
I usually turn on the PC, and pop back a minute later, without autofsck I would have missed the chance to escape the disk check, and will be unable to use the PC for 15 minutes.

Also with hardy a disk check will be required at some point. I prefer disk checks on shutdown where I don't need to be sitting around waiting. I leave it checking knowing it will power down afterwards.]]> Sun, 02 Mar 2008 17:16:49 +0000
Some have suggested that the check happen at shutdown instead. I can't say I'm fond of that idea unless I can opt to cancel that as well. I'm running on a laptop. Often when I want to shutdown I need to shutdown NOW.

The best idea I've heard is that fsck be made safe to run on a mounted disk so that it can run in the background.

]]> Sun, 02 Mar 2008 22:21:00 +0000 I don't know if it can be fixed, but the golden rule right now is never, never, NEVER run an fsck on a mounted file system! You will almost certainly regret what happens to it. Corrupted partition, lots of files moved to lost & found, heavy general wreckage and other cataclims alike.

The exception to this is if the file system is mounted read-only, in which case it is safe to do so. But then, running read-only means any program that needs to write to disk is out of luck. I don't know if it affects fifo files (pipes), but anything that needs read-write access to /var/spool or some other directory may find it's startup jeopardized. ]]> Mon, 03 Mar 2008 23:29:37 +0000 I have 2 harddisk and when it checks my slave one, it just block here, saying that this drive was not checked since 20 times...
I have to restart on a live CD and run a fsck command by myself every 20 days...

quite bothering ;)
]]> Wed, 05 Mar 2008 09:46:56 +0000
I just think this should be handled in a more transparent option. Would it be possible to add a visual indicator to GRUB that tells you how many times left until a "compulsory" fsck, and include an option to do it early or late?]]> Thu, 06 Mar 2008 16:10:35 +0000
It's too bad that this fine technology is being overlooked because of social problems of the original programmer. It used to be the default of many distros.]]> Sat, 08 Mar 2008 14:01:34 +0000 Sat, 08 Mar 2008 23:20:27 +0000 Sun, 09 Mar 2008 17:44:07 +0000 Mon, 10 Mar 2008 15:33:40 +0000 Tue, 11 Mar 2008 16:00:53 +0000
You could use EXT3 without a fsck very well and I think that it sheds a wrong light on it (makes it look inferior to NTFS).
It should definitely be interuptible, pointing out to the user that it is recommended to do a routine maintenance once in while to make sure you data is safe and sound and ordered.
(most users don't understand the severity and think that it is something serious (which it is not)


]]> Thu, 13 Mar 2008 02:03:18 +0000 I am on kubuntu 8.04 amd64 and I still cannot skip fsck at boot time.]]> Sat, 29 Mar 2008 22:57:21 +0000
http://blog.linuxoss.com/2006/09/27/suse-102-ditching-reiserfs-as-it-default-fs/]]> Tue, 29 Apr 2008 11:48:18 +0000
surely by default this should be done on shutdown, or in the background, and not on boot-up.]]> Thu, 01 May 2008 06:07:45 +0000 Ext3 is a filesystem, who inherited all the crappy by his predecessor, with the only advantage of journal feature (mandatory, because ext2 was loosing data seriously).
You can tune the fsck request by using tune2fs, but, anyway, have you never seen a filesystem requiring FSCK after certain amount of mount operations?
Does Veritas, ZFS, UFS, XFS, Reiser, or even your windows XP, or windows Server, or JFS on hp-ux, or ibm-jfs on linux, requires to run FSCK after mounting the disk many times?
Furthermore, is there any relation between the number of mounting and the filesystem usage?
A huge fileserver, serving thousand of users, which was never reboot, will FSCK every 180 day, a desktop rebooted twice a day, will run FSCK every 20 days.
But if you never reboot such server, you'll never fsck on it.
Does this mean that you have to reboot a machine with ext3, at least, every 180 day (even if it's not written anywhere)?

Furthermore, I had problems with ext3 few times, but I never had problems with any other filesystem (no problems with VXFS, not problems with UFS, no problems with Reiser, no problems with XFS, and no problems with my windows with NTFS).
The only FS which can guarantee you having problem is EXT3.

I would suggest to NOT use ext3.
I don't have experience with JFS on linux, but I can ensure that Reiser and XFS work properly.

p.s.: XFS (like VXFS from Veritas) has also defrag tool (called xfs_fsr) because journaled FS have fragmentation problems. So... I can't find one reason to use ext3.
]]> Fri, 09 May 2008 19:48:59 +0000
esc, ctrl-c, spacebar, return, enter, smashing the keyboard in frustration do absolutely nothing except print pretty lines on the screen. ]]> Mon, 19 May 2008 05:09:23 +0000
It's bad that it doesn't write that by default. See http://brainstorm.ubuntu.com/idea/13302/ and vote.
]]> Tue, 16 Sep 2008 14:54:54 +0000