Hello, On Tue, 20 Dec 2011, Felix Miata wrote:

On 2011/12/20 19:26 (GMT) Thomas Hertweck composed:

...

On 20/12/11 18:59, Istvan Gabor wrote:

...

At every Nth mount fsck checks a given partition, which takes a long time. Sometimes I do need the computer quickly and don't want to wait until fsck finishes. Is there an option that makes possible to interrupt or cancel file system checking during boot so that the checking would be completed at the next boot? I don't want to disable it or change the max count, I just want occasionally bypass checking at boot. I tried to interrupt fsck by pressing ctrl+c but it resulted in a not normal boot, many partitions were not mounted.

...

Append "fastboot" as boot parameter when the grub menu shows up. [..] I set all my partitions to either 12m or 0 time count, and 0 mount count, with tune2fs, and either let the init mount process decide if and when a filesystem needs to be repaired before mounting, or delay automatic fsck at least a year.

I set all partitions but the /-partition to "not check" in fstab (0 in last column), and have varying max mount counts. As I have no splash, I see that a FS should be checked (at least ext3 tells you so: [ 232.782545] EXT3-fs (sda2): warning: maximal mount count reached, running e2fsck is recommended ), and I also have a script to check: ,----[ /root/bin/check_mountcount.sh ] | #!/bin/sh | printf "%-10s %4s %4s %s\n" "device" "max" "cur" "mounted on" | echo "---------- ---- ---- ----------···" | | grep 'ext[234]' /proc/mounts | while read device mntpt rest | do | tune2fs -l "$device" 2>/dev/null | \ | awk '/^Mount count/ { cur=$3; } | /^Max.*mount count/{ max=$4; } | END { printf "%-10s %4s %4s %s\n", dev, max, cur, mnt; }' \ | dev="$device" mnt="$mntpt" - | done `---- and if I don't need a partition a while (and little disk anyway), I unmount that partition and run fsck on it. HTH, -dnh -- That's O'Neill with two L's, the other one has no sense of humor. -- Col. Jack O'Neill, Stargate -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org

2011. december 20. 21:46 napon Anton Aylward írta:

Istvan Gabor said the following on 12/20/2011 01:59 PM:

...

At every Nth mount fsck checks a given partition, which takes a long time.

fsck is ¬* linear with the size of a file system but a function of the power (square) or the size.

So a 4G FS will take longer to process than two 2G FS. Also the 2 file systems will be able to the fsck'd in parallel :-) You might also consider a setup where the 2 FS are _not_ fsck'd together: both have the same N but are spaced apart by M.

Thank you all for your responses. From the answers it seems I haven't explained clearly what I wanted. I don't have any problem with file system checking at boot, I think it is good. Therefore I don't want to change the change max-mount-counts (number). My problem is that occasionally I do not have the time to wait until fsck stops checking even one partition (eg is takes 5 minutes but I need to use the system within a minute). As I am not following the partitions' mount counts I don't know in advance whether a partition will be checked or not, when I start the system. When I see that fsck already started to check the filesystem I'd like to press a button (eg escape) which signals to the system that I want to cancel or interrupt checking and continue with the next step in the boot process. Is it possible? If not, it should be implemented. Thomas, is the 'fastboot' option you mentioned related to the "doing fast boot" message shown at the beginning of boot? Thanks, Istvan -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org

On 20/12/11 21:51, Istvan Gabor wrote:

Thomas, is the 'fastboot' option you mentioned related to the "doing fast boot" message shown at the beginning of boot?

Please check /etc/init.d/boot and /etc/init.d/boot.rootfsck or /etc/init.d/boot.localfs, respectively, if you are interested in the details. These are the names of the files on my 11.4 but 11.2 will probably be identical. ----> /etc/init.d/boot [...] DO_FASTBOOT=no [...] if rc_cmdline fastboot > /dev/null || test -e /fastboot ; then DO_FASTBOOT=yes fi [...] This means, if the kernel command line contains a "fastboot" parameter or if there exists a file /fastboot in the root filesystem, then no regular filesystem check will be performed for this particular system startup as you can see in the other two boot scripts. The file /fastboot will be removed by boot.localfs, i.e. the fsck will only be skipped for the current startup: [...] # # clean up # rm -f /etc/nologin /nologin /fastboot /forcefsck /succes [...] You can check the used kernel command line via "cat /proc/cmdline" once the system is up. As far as I can see, this is exactly the feature you were looking for. Just add "fastboot" to the kernel command line when you need it and regular filesystem checks will be skipped. Thomas -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org

2011. december 21. 0:46 napon "Brian K. White" írta: [snip]

I think your best option is to:

* make sure all large fs's are journaling (reiserfs, ext4, etc)

* drastically increase or disable altogether the periodic boot time scanning (not disabling the fsk if the fs is found to be dirty at boot)

* manually fsck once in a while if you like.

Because you can't predict when you will want to boot and run as fast as possible, and those times are _exactly_ the times you _don't_ want to have to futz with special boot prompt parameters.

You could also make a menu.lst stanza that just has fastboot added to it.

That way you could boot up fast without having to stop and enter special boot prompt options. Just select the fastboot entry instead of the regular entry.

fsck does process interrupt signals gracefully. You could possibly modify the initrc or other boot script to allow a break signal (ctrl-c) to reach the fsck process (it's probably trapped by a trap command before fsck runs normally, but you could edit the script with another trap command that releases that trap just before fsck). And aborting a read-only checkup fsck should always be safe.

But aborting fsck at all is just a really unwise habit to give yourself.

So I would do one of the other things that avoids running fsck in the first place.

Thank you again. I guess I should increase max mount count and also make a fastboot entry in menu.lst. If I know before boot that I need the system urgently I use fastboot, otherwise the regular boot with increased max mount numbers. Following the drives' states (actual mount count) is not an option as I have ~15 partitions, some mounted automatically, some by users if necessary. Thanks again, Istvan -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org

On 12/23/2011 4:27 PM, Carlos E. R. wrote:

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1

On 2011-12-24 00:26, Istvan Gabor wrote:

...

Following the drives' states (actual mount count) is not an option as I have ~15 partitions, some mounted automatically, some by users if necessary.

It would if you can track the mount count and the maximal count, so as to predict “next time will fsck one drive”. If this is checked by an script on boot, it might help.

Good idea. Why not check this on shutdown, and offer the ability to fsck and shutdown when completed? After all, you are not planning to use the computer when you tell it to shut down and you can be safely logged out, the partition dismounted and checked without inconveniencing the user. -- _____________________________________ ---This space for rent--- -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org

Le 24/12/2011 07:40, Jon Clausen a écrit :

These days I rarely use 'ext' filesystems for anything but /boot/, but I've walked into the "unexpected fsck" enough times that I can see some value in at least being forewarned.

do you have any number on the fsck time? I just has a checkout of my desktop (2 drives, 1 and 2 Tb) and it was less than one minute (ext4), so the waiting time is manageable. (openSUSE 12.1) I think having it at shutdown is worst, because, I often shutdown in a hurry (not having time to look at messages). But a waiting time with ability to disable the control at boot could be a good solution, if I boot in a hurry, usually I'm in front of the computer, waiting and I can answer sorry if all this was already answered, I couldn't follow all the thread jdd -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org

On Saturday 24 December 2011 03:36:05 John Andersen wrote:

On 12/23/2011 4:27 PM, Carlos E. R. wrote:

...

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1

On 2011-12-24 00:26, Istvan Gabor wrote:

...

Following the drives' states (actual mount count) is not an option as I have ~15 partitions, some mounted automatically, some by users if necessary.

It would if you can track the mount count and the maximal count, so as to predict “next time will fsck one drive”. If this is checked by an script on boot, it might help.

Good idea.

Why not check this on shutdown, and offer the ability to fsck and shutdown when completed?

But please only actively forced by admin doing something like shutdown --fsck-after-umount Usually I don't care about bootup time but shutdown has to be as fast as possible per default.

After all, you are not planning to use the computer when you tell it to shut down and you can be safely logged out, the partition dismounted and checked without inconveniencing the user.

Adding a shutdown option --fsck-after-umount and making shutdown -f (Skip fsck on reboot) the default would IMO solve the whole problem. cu, Rudi -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org

Hello, On Wed, 21 Dec 2011, John Andersen wrote:

On 12/20/2011 8:43 PM, sc wrote:

...

all it takes is a little attention -- if, for you, as it does for me now, fsck starts a 60 day timer when it performs its chore,

Does it not also count boot ups? It used to do that as well, but maybe it has dropped that behavior in recent years, I haven't been paying attention.

The OP is shutting down and rebooting every day, not suspend to ram, but a full windows user style reboot every morning.

# tune2fs -l /dev/root ### [pruned] Mount count: 1 Maximum mount count: 36 Check interval: 15552000 (6 months) Next check after: Mon Jun 18 14:59:25 2012 # tune2fs -l /dev/sdi1 ### [pruned] Mount count: 854 Maximum mount count: 75 Last checked: Thu Feb 25 14:42:31 2010 Check interval: 15552000 (6 months) # ls -l /dev/disk/by-label/OLDHDC [..] /dev/disk/by-label/OLDHDC -> ../../sdi1 # grep ' / \|hdc' /etc/fstab LABEL=P_SUSE / ext3 [options] 1 1 LABEL=OLDHDC /mnt/hdc1 ext3,ext2 [options] 0 0 I think I really should check sdi/hdc sometime again, that the 500G IDE disk out of the old box where it _was_ hdc for years. ;) # smartctl -A /dev/sdi1 | awk '$1 == 9 { print; }' 9 Power_On_Hours [..] 11715 HTH, -dnh -- Warning: Pregnancy can cause birth -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org