View previous topic :: View next topic |
Author |
Message |
vincent_bachelier Tux's lil' helper
Joined: 26 Sep 2004 Posts: 96
|
Posted: Tue Dec 13, 2005 9:36 pm Post subject: |
|
|
I already use noatime ... I will try nodiratime !
I use nitro sources with ck patches inside ... mm correction are certainly not include in it
but mm are a usable kernel ? _________________ Shaolan Oji
Blog : http://ilwg.blogspot.com [ Gentoo Fan Blog Site ]
Project:
vbwallpapers: http://vbwallpapers.berlios.de [ get random wallpapers for your desktop ] |
|
Back to top |
|
|
gimpel Advocate
Joined: 15 Oct 2004 Posts: 2720 Location: Munich, Bavaria
|
Posted: Tue Dec 13, 2005 9:41 pm Post subject: |
|
|
vincent_bachelier wrote: | I already use noatime ... I will try nodiratime !
I use nitro sources with ck patches inside ... mm correction are certainly not include in it
but mm are a usable kernel ? |
i don't know what nitro's reiser4 version is...
but -mm is really not a good choice! very unstable.. i just use the reiser4 patches from -mm and add those to -ck etc _________________ http://proaudio.tuxfamily.org/wiki - pro-audio software overlay
|
|
Back to top |
|
|
infirit l33t
Joined: 11 Jan 2003 Posts: 778 Location: Hoofddorp / The Netherlands
|
Posted: Tue Dec 13, 2005 10:18 pm Post subject: |
|
|
gimpel wrote: | infirit wrote: | Reiser4 is currently horribly unstable (after 2.6.13) and because of this I switched back to reiserfs. Hanging on high load and fsync performance are 2 of the know reiser4 issues.
Also there is no resizer yet so people using lvm2 (like me) can not grow or shrink partitions. |
all fixed in -mm already.
you could also mount reiser4 with noatime,nodiratime to 'solve' those issues in 2.6.14
and resizing is evil anyway | Afaik there are no new reiser4 patches since 2.6.14-rc5-mm1 and that one still has the problems described. And how would resizing be evil _________________ EASY TO INSTALL = Difficult to install, but instruction manual has pictures.
Join the adopt an unanswered post initiative today |
|
Back to top |
|
|
vincent_bachelier Tux's lil' helper
Joined: 26 Sep 2004 Posts: 96
|
Posted: Wed Dec 14, 2005 6:47 am Post subject: |
|
|
Well, thx to noadirtime and noatime, now performance are really good, and no hang up anymore ...
I hope this problem will be solve in a sooner futur and reiser4 include in vanilla sources ...
my wrapper aren't usefull because of the -a option ... I think they rarelly need to be check because they keep a concistant state every time
even my battery go down .. !!!!
I will try this a month and see ...
bye _________________ Shaolan Oji
Blog : http://ilwg.blogspot.com [ Gentoo Fan Blog Site ]
Project:
vbwallpapers: http://vbwallpapers.berlios.de [ get random wallpapers for your desktop ] |
|
Back to top |
|
|
vincent_bachelier Tux's lil' helper
Joined: 26 Sep 2004 Posts: 96
|
Posted: Thu Dec 15, 2005 8:39 am Post subject: |
|
|
Well, I have improbe my wrapper, now work with forcefsck file option ...
here the code:
Code: |
#!/bin/sh
PARAMS="--fix $@"
[ -e /forcefsck ] && PARAMS=${PARAMS/ -a/}
echo "Reiser4 check with \"$PARAMS\" ..."
exec $0.orig $PARAMS
|
it filter -a option when forcefsck is need ...
then I have make a /etc/init.d/forcefsck script to active check every 100 boottime
Code: |
#!/sbin/runscript
BOOT_COUNT="100"
[ -e /etc/conf.d/forcefsck ] && source /etc/conf.d/forcefsck
[ -e /etc/forcefsck.boot ] || echo 0 > /etc/forcefsck.boot
BOOT_CURRENT=`cat /etc/forcefsck.boot`
BOOT_CURRENT=`expr $BOOT_CURRENT + 1`
echo $BOOT_CURRENT > /etc/forcefsck.boot
BOOT_STAY=`expr $BOOT_COUNT - $BOOT_CURRENT`
start() {
ebegin "ForceFsck in $BOOT_STAY boottime ..."
if [ $BOOT_STAY -le 0 ]
then
touch /forcefsck
echo 0 > /etc/forcefsck.boot
fi
eend 0
}
|
Now it check sometime all filesystem like ext2/ext3 do
for the moment I lose one file, a library, re-emerge the package solve the problem ...
the tail mode, put small file into hash branches, isn't it ?
then is more secure for don't lose small file when hash table crash to put notail option no ?
what do you think ? _________________ Shaolan Oji
Blog : http://ilwg.blogspot.com [ Gentoo Fan Blog Site ]
Project:
vbwallpapers: http://vbwallpapers.berlios.de [ get random wallpapers for your desktop ] |
|
Back to top |
|
|
Gentree Watchman
Joined: 01 Jul 2003 Posts: 5350 Location: France, Old Europe
|
Posted: Tue Jan 17, 2006 6:09 pm Post subject: |
|
|
@vincent_bachelier
just to confirm one of the posts above , the whole design of reiserfs and reiser4 is based on atomic writes, this should mean you are covered in the case of a power out. You do not need to fsck on boot so set the last param to zero as fstab as suggested.
This is one of the big advantages of R4 and I have been saved on several occations when the local power supply cut out.
you should only fsck.reiser4 when you have a specific problem. IF its your root partition do this from a liveCD.
BIG WARNING . There is an incompatability between the kerenl support for 1.0.4 and 1.0.5 (note I am not refering to reiser4progs here.)
This generally means that if you mount (and I mean just mount) an R4 on a later kernel - >=2.6.12 it will _apear_ corrupted if you attempt to use it with an earlier kernel , like the one you might find on a R4 liveCD!
I've posted more detail elsewhere if you're affected. Generally if you know of the issue , you will not need to deal with the consequences.
I breifly dabbled with xfs on a test partition but lost interest when I had a corruption that lost the whole lot. Perfomance is hottly discussed but it can be very fast in certain very limiitted conditions but overall is a fair bit slower than R4 and ext3.
Despite what some may say the tools do generally work. I successfully recovered from the above incompatability prob and managed to retro to 1.0.4 using fsck.rieser4 . I critisise reiser for not even forseeing this prob but the tools did work.
Also be very careful not to run out of space on a R4 root partition. ext3 reserves 5% for the system to ensure it can always function, R4 does not and things get very unstable when the system cannot even create a tmp file or anything. Luckily you can power off an R4 system in this case but it's not too tidy.
An R4 that runs out of space often causes a corruption. I have just fixed this on another partition where portage_tmpdir overfilled the available space. Again the tools did the job without any fuss.
happy reisling. _________________ Linux, because I'd rather own a free OS than steal one that's not worth paying for.
Gentoo because I'm a masochist
AthlonXP-M on A7N8X. Portage ~x86 |
|
Back to top |
|
|
plasmagunman l33t
Joined: 07 Jun 2002 Posts: 604 Location: berlin
|
Posted: Tue Jan 17, 2006 6:30 pm Post subject: |
|
|
Gentree wrote: | just to confirm one of the posts above , the whole design of reiserfs and reiser4 is based on atomic writes, this should mean you are covered in the case of a power out. You do not need to fsck on boot so set the last param to zero as fstab as suggested. |
that's wrong imho: reiser4 has atomic write, reiserfs does not. reiserfs is just journaled, and the idea of a journal is to find the unfinished writes on disk faster. so in case of a crash you don't have to scan the entire disk, just the part where the last unfinished write was happening.
and there's still the possibility of a corrupted filesystem caused by hardware-problems or buggy filesystem code. supposing that the code is not 100% bug free you have to fsck once in a while. i do that with a manual "shutdown -r -F now", check this thread for details.
Quote: | Also be very careful not to run out of space on a R4 root partition. ext3 reserves 5% for the system to ensure it can always function, R4 does not and things get very unstable when the system cannot even create a tmp file or anything. |
same problem exists with reiserfs.
sorry, i'm in a hurry, i explained my thoughts about fsck in another thread, when i had more time... _________________ please, feel free to correct my english. - por favor, corrige mi español. |
|
Back to top |
|
|
Gentree Watchman
Joined: 01 Jul 2003 Posts: 5350 Location: France, Old Europe
|
Posted: Tue Jan 17, 2006 9:14 pm Post subject: |
|
|
Thanks for your comments and the other post.
shame Hans Reiser did not use the same option syntax as existing fs , but I guess that's not his style
I do a manual check from time to time but I find R4 so solid that it is pretty rarely needed.
you are correct only R4 uses atomic writes (at least is does unless the disk is full in which case it seems it can split the atom!)
Quote: | same problem exists with reiserfs |
I have not checked this but I would not mind betting this applies everyting except ext2/3 .
R4 then remains the fs the most likely to recover from this situation without major corruption.
ext3 would likely recover poorly that's why the system reserve is there. This leads to possibility of recovering a nice chunk of space if you use ext3. see -m option.
Modern disks having jumped in capacity you may find you 5% is around 10GB for a 200G drive! I suspect a few hundred megs would do the job.
man mkfs.ext3 wrote: |
NAME
mke2fs - create an ext2/ext3 filesystem
SYNOPSIS
mke2fs [ -c | -l filename ] [ -b block-size ] [ -f fragment-size ] [ -g
blocks-per-group ] [ -i bytes-per-inode ] [ -j ] [ -J journal-options ]
[ -N number-of-inodes ] [ -n ] [ -m reserved-blocks-percentage ] [ -o
creator-os ] [ -O feature[,...] ] [ -q ] [ -r fs-revision-level ] [ -E
extended-options ] [ -v ] [ -F ] [ -L volume-label ] [ -M last-mounted-
directory ] [ -S ] [ -T filesystem-type ] [ -V ] device [ blocks-count ]
|
_________________ Linux, because I'd rather own a free OS than steal one that's not worth paying for.
Gentoo because I'm a masochist
AthlonXP-M on A7N8X. Portage ~x86 |
|
Back to top |
|
|
|