1 May
2003
1 May
'03
9:31 p.m.
dear list,
i'm running Debian kernel version 2.4.20 with ext3 an an Asus L3 laptop
with a Hammerfall HDSP, and my xruns using Jack are frequent and nasty!
i would say in the nieghborhood of every 10-20 seconds, i get an xrun of
10-200 ms. i have done all the hdparm stuff, all the necessary low
latency and preemptive patches, and have everything set up to run
realtime capabilities for Jack. the problem, of course, is that with
ext3 and kernel 2.4.20, the machine totally freezes when i do a
"jackstart --realtime". [this is documented somewhere, so i know it's
not just me!]
so, i figure my solution is to convert the filesystem of my main
partition to reiserfs, thereby allowing me to gain from its better
journal [i.e. less disk activity, i guess] and the ability to use
realtime capabilities.
and here is where i start to feel like a dumb newbie :-)
if my main partition is /dev/hda2, and i am working from another
[temporary] root file system in /dev/hda6, how can i back up /dev/hda2
to /dev/hda7 and put it back again after doing the ext3-->reiserfs
conversion WITHOUT transferring back over the old file system? i tried
dd if=/dev/hda2 of=/dev/hda7, but when i mounted /dev/hda7, it was
turned into an ext3 filesystem, although i had already made it reiserfs.
i assume that means if i changed dev/hda2 to reiserfs and moved the
backup copy there again, it would be ext3 all over again. i tried a few
things with cpio that didn't work, and i am not sure i have the syntax
right with tar. can someone please help me out here?
thx.
derek
1 May
1 May
10:03 p.m.
On Fri, May 02, 2003 at 12:53:28AM +0200, derek holzer wrote:
conversion WITHOUT transferring back over the old file
system? i tried
dd if=/dev/hda2 of=/dev/hda7, but when i mounted /dev/hda7, it was
turned into an ext3 filesystem, although i had already made it reiserfs.
yep. dd does raw byte copying; you successfully wiped out your new
reiserfs and dumped a binary image of your old fs onto it :)
i THINK you can just do this (assuming all partitions are mounted)
make the backup:
tar -czvf /where/hda7/is/mounted/backup.tgz /where/hda2/is/mounted/
then create the reiserfs partition on /dev/hda2 (assuming you're not
booted from there)
mount /dev/hda2 somewhere
cd somewhere
tar -xzvf /where/hda7/is/mounted/backup.tgz
then if /dev/hda2 is your root partition (i'm not clear from what you
said) edit somewhere/etc/fstab appropriately
and if you /boot is on /dev/hda2, you'll need to run lilo
telling it to use somewhere/etc/lilo.conf (i forget what flag you
use ... man lilo) ... unless of course you don't use lilo :)
sorry for not being more clear, i haven't done this in a year or two.
--
Paul Winkler
home: http://www.slinkp.com
"Muppet Labs, where the future is made - today!"
10:10 p.m.
On Fri, 2 May 2003 08:53 am, derek holzer wrote:
if my main partition is /dev/hda2, and i am working
from another
[temporary] root file system in /dev/hda6, how can i back up /dev/hda2
to /dev/hda7 and put it back again after doing the ext3-->reiserfs
conversion WITHOUT transferring back over the old file system? i tried
dd if=/dev/hda2 of=/dev/hda7, but when i mounted /dev/hda7, it was
turned into an ext3 filesystem, although i had already made it reiserfs.
i assume that means if i changed dev/hda2 to reiserfs and moved the
backup copy there again, it would be ext3 all over again. i tried a few
things with cpio that didn't work, and i am not sure i have the syntax
right with tar. can someone please help me out here?
Something like this?
mkdir /mnt/part1
mkdir /mnt/part2
mount -t ext3 /dev/hda2 /mnt/part1
mount -t ext3 /dev/hda7 /mnt/part2
cp -av /mnt/part1 /mnt/part2
[copious output]
umount /mnt/part1
mkreiserfs /dev/hda2
mount -t reiserfs /dev/hda2 /mnt/part2 -o noatime,notail
cp -av /mnt/part2 /mnt/part1
[copious output]
umount /mnt/part1
umount /mnt/part2
rmdir /mnt/part1
rmdir /mnt/part2
vi /etc/fstab
reboot
With my own system, I have seperate partitions for /home, /tmp, /usr, /var,
/boot, and /, all previously in ext3. I changed fstab to mount these all as
ext2 instead, and converted my /home to reiserfs using the process above (or
close to). Ardour runs much nicer now.
t
--
GPG : http://n12turbo.com/tarragon/public.key
11:07 p.m.
On Fri, 2 May 2003 09:56 am, Tarragon Allen wrote:
On Fri, 2 May 2003 08:53 am, derek holzer wrote:
Something like this?
mkdir /mnt/part1
mkdir /mnt/part2
mount -t ext3 /dev/hda2 /mnt/part1
mount -t ext3 /dev/hda7 /mnt/part2
cp -av /mnt/part1 /mnt/part2
[copious output]
umount /mnt/part1
mkreiserfs /dev/hda2
mount -t reiserfs /dev/hda2 /mnt/part2 -o noatime,notail
Typo. That should be (of course):
mount -t reiserfs /dev/hda2 /mnt/part1 -o noatime,notail
cp -av /mnt/part2 /mnt/part1
[copious output]
umount /mnt/part1
umount /mnt/part2
rmdir /mnt/part1
rmdir /mnt/part2
vi /etc/fstab
reboot
--
GPG : http://n12turbo.com/tarragon/public.key
2 May
2 May
4:37 a.m.
Hallo,
Tarragon Allen hat gesagt: // Tarragon Allen wrote:
Something like this?
[...]
mount -t ext3 /dev/hda7 /mnt/part2
cp -av /mnt/part1 /mnt/part2
[...]
cp -av /mnt/part2 /mnt/part1
Make VERY SURE, that you use the -x switch:
cp -avx
to not copy things from other partitions, that are symlinked.
Otherwise this is also my solution: It is equivalent to the tar-pipe,
whose syntax I never remembered...
ciao
--
Frank Barknecht _ ______footils.org__
5:18 a.m.
On Fri, 2 May 2003 04:26 pm, Frank Barknecht wrote:
Hallo,
Tarragon Allen hat gesagt: // Tarragon Allen wrote:
I thought that would be handled by the -a, which basically turns on -dpR ..
the -d being the important one - "same as --no-dereference --preserve=link"
..
t
--
GPG : http://n12turbo.com/tarragon/public.key
Something like this?
[...]
mount -t ext3 /dev/hda7 /mnt/part2
cp -av /mnt/part1 /mnt/part2
[...]
cp -av /mnt/part2 /mnt/part1
Make VERY SURE, that you use the -x switch:
cp -avx
to not copy things from other partitions, that are symlinked.
Otherwise this is also my solution: It is equivalent to the tar-pipe,
whose syntax I never remembered...
ciao
11:53 a.m.
Hallo,
Tarragon Allen hat gesagt: // Tarragon Allen wrote:
Uhm, yep, my explanation was wrong:
-x, --one-file-system
stay on this file system
means: don't copy direcories, that are mountpoints....
ciao
--
Frank Barknecht _ ______footils.org__
Make VERY SURE, that you use the -x switch:
cp -avx
to not copy things from other partitions, that are symlinked.
Otherwise this is also my solution: It is equivalent to the tar-pipe,
whose syntax I never remembered...
ciao
I thought that would be handled by the -a, which basically turns on -dpR ..
the -d being the important one - "same as --no-dereference --preserve=link"
..
12:12 p.m.
On Fri 02/05/2003 16:46:30, Tarragon Allen wrote:
On Fri, 2 May 2003 04:26 pm, Frank Barknecht wrote:
-x, --one-file-system
stay on this file system
Which is different from not following links. This prevents filesystems
mounted within the one to be copied from being copied.
Hallo,
Tarragon Allen hat gesagt: // Tarragon Allen wrote:
I thought that would be handled by the -a, which basically turns on -dpR ..
the -d being the important one - "same as --no-dereference --preserve=link"
.. Something like this?
[...]
mount -t ext3 /dev/hda7 /mnt/part2
cp -av /mnt/part1 /mnt/part2
[...]
cp -av /mnt/part2 /mnt/part1
Make VERY SURE, that you use the -x switch:
cp -avx
to not copy things from other partitions, that are symlinked.
Otherwise this is also my solution: It is equivalent to the tar-pipe,
whose syntax I never remembered...
ciao
t
--
GPG : http://n12turbo.com/tarragon/public.key
7:53 a.m.
thanks everybody for the syntax :-)
one further question, regarding this:
Tarragon Allen wrote:
With my own system, I have seperate partitions for
/home, /tmp, /usr, /var,
/boot, and /, all previously in ext3. I changed fstab to mount these all as
ext2 instead, and converted my /home to reiserfs using the process above (or
close to). Ardour runs much nicer now.
what part of my file system benefits the most from running reiserfs?
right now, i have everything on one partition, but if i were to break
this up with seperate /home and / partitions, should they both be
reiserfs? in short, is it the operating system that benefits from the
reiserfs, or the directory where your sound files live? i haven't done
any recording at all yet due to my bad xruns, so it would seem [thus
far] that my root partition needs a faster file system. or would it be
better to use a non-journalling system for the root partition [as you
have above] and reiserfs for the /home partition where the sound gets
streamed to and from?
also, would one use a similar proceedure to the one you have described
to move the contents of a /home directory in the root partition to a
seperate /home partition?
thx + best,
derek
1:43 p.m.
On Fri, 2 May 2003 07:15 pm, derek holzer wrote:
what part of my file system benefits the most from
running reiserfs?
right now, i have everything on one partition, but if i were to break
this up with seperate /home and / partitions, should they both be
reiserfs? in short, is it the operating system that benefits from the
reiserfs, or the directory where your sound files live? i haven't done
any recording at all yet due to my bad xruns, so it would seem [thus
far] that my root partition needs a faster file system. or would it be
better to use a non-journalling system for the root partition [as you
have above] and reiserfs for the /home partition where the sound gets
streamed to and from?
I'm partial to using journalling filesystems so that if the system hangs I'm
less likely to lose data, and reboots will be a lot faster. This is the main
reason I switched to ext3.
Reiserfs is apparently the filesystem of choice if you want a high performance
journalling filesystem that will handle many small files (ie: databases,
squid caches, etc). In my case, I would have just as happily stuck with ext3
except that Ardour/Jack managed to lock my system every time until I changed
(even with a patched kernel 2.4.20).
I would say for very small partitions, there's not much use using journalling
as the journal itself takes up space and for small partitions the reboot
auto-fsck doesn't take that long anyway. Otherwise most things will benefit
from journalling, but only in so much as it's a safety precaution. I'm not so
sure that there's a performance benefit from any journalling filesystem, as
by definition it's doing more than a non-journalling filesystem, and so there
is extra overhead involved.
Reiserfs is highly recommended by the Gentoo crowd, if that makes any
difference.
also, would one use a similar proceedure to the one
you have described
to move the contents of a /home directory in the root partition to a
seperate /home partition?
Yes. Just cd in /home and do the "cp -axv" (yes, that's an updated version!
:)
to your new partition. The main reason for seperating partitions is so that
if one program goes crazy and writes lots and lots of data to disk it's not
going to kill your system. This is particularly important for /var, where all
the logs go, but can be important in /home. Another reason for seperating
partitions is so that you can easily re-install or try other distributions,
yet keep some of the same data between installs/versions (/home being the big
example).
I have /boot seperate for this reason as well - I can have a gentoo kernel and
a debian kernel in there, and swap between them happily yet use the same old
/boot partition.
t
--
GPG : http://n12turbo.com/tarragon/public.key
12:17 p.m.
dear list,
i'm running Debian kernel version 2.4.20 with ext3 an an Asus L3 laptop
with a Hammerfall HDSP, and my xruns using Jack are frequent and nasty!
i would say in the nieghborhood of every 10-20 seconds, i get an xrun of
10-200 ms.
<SNIP>
Various emails about converting to reiserfs...
<SNIP>
If the idea here is that the xruns are caused by the ext3 journaling
operations, wouldn't a simple test be to mount the ext3 partition as ext2
and prove it before doing all the work to move to reiserfs only to find out
it's something else? Or possibly change the journaling commit time to reduce
or increase the xruns?
Maybe it's too late for that. Sorry...
12:56 p.m.
On Fri, May 02, 2003 at 07:06:31AM -0700, Mark Knecht wrote:
If the idea here is that the xruns are caused by the
ext3 journaling
operations, wouldn't a simple test be to mount the ext3 partition as ext2
and prove it before doing all the work to move to reiserfs only to find out
it's something else?
i'd think so!
--
Paul Winkler
home: http://www.slinkp.com
"Muppet Labs, where the future is made - today!"
1:40 p.m.
On Sat, 3 May 2003 12:06 am, Mark Knecht wrote:
I think it was more to do with outright lockups using the ext3 code, rather
than xruns. Yes, switching to ext2 is a good way to test if ext3 is at fault
(I did this, and yes it was). I'm using reiserfs as a workaround (I'd prefer
to have any journalling on my larger filesystems than none at all), although
from what I've read reiserfs might be better suited to audio applications
anyway.
t
--
GPG : http://n12turbo.com/tarragon/public.key
dear list,
i'm running Debian kernel version 2.4.20 with ext3 an an Asus L3 laptop
with a Hammerfall HDSP, and my xruns using Jack are frequent and nasty!
i would say in the nieghborhood of every 10-20 seconds, i get an xrun of
10-200 ms.
<SNIP>
Various emails about converting to reiserfs...
<SNIP>
If the idea here is that the xruns are caused by the ext3 journaling
operations, wouldn't a simple test be to mount the ext3 partition as ext2
and prove it before doing all the work to move to reiserfs only to find out
it's something else? Or possibly change the journaling commit time to
reduce or increase the xruns?
Maybe it's too late for that. Sorry... 
2:04 p.m.
I'm using reiserfs as a workaround (I'd
prefer
to have any journalling on my larger filesystems than none at all),
although from what I've read reiserfs might be better suited to
audio applications anyway.
I've been using it for a long time without problems. I use it on all
partitions except /boot and swap.
I usually convert a machine to ReiserFS during Mandrake installs -
after a full backup of /home of course. The graphical partition tool
makes this very easy.
Cheers
Daniel
2:31 p.m.
I'm using reiserfs as a workaround (I'd prefer
to have any journalling on
my larger filesystems than none at all), although from
what I've read
reiserfs might be better suited to audio applications anyway.
t
I guess this also has to do with the way you've built a specific machine. In
my case, I place all of my audio on a completely separate 1394 hard drive,
and that 1394 drive is reiserfs. I make backups of the audio data to an
older 1394 drive that's FAT32, which allows me to easily move audio to my
Windows platform (where I do most of my work) without having to deal with
Samba or ftp. (How do others move data between Windows and Linux?)
My system disk is ext3 simply because it's more tested than reiserfs, so
possibly more reliable. When I'm doing audio work, to the best of my
knowledge, there are no transactions taking place between the computer and
the ext3 drive anyway. (i.e. - I see activity lights on my 1394 reiserfs
drive, but none on my internal EIDE ext3 drive.)
Now that I'm starting to use Gentoo, I can more easily try out XFS on my
1394 audio drive which should be better for large file sizes.
I feel like this setup is giving me pretty much the best of all
capabilities, but I'm a newbie in Linux and certainly open to more ideas.
I'm not sure journaling is even working in my case as fsck's are still
pretty slow the couple of times that has happened.
- Mark
3:12 p.m.
How do others move data between Windows and Linux?
FTP over wireless LAN. It's very easy with the likes of the KDE file
manager - just drag and drop.
My system disk is ext3 simply because it's more
tested than
reiserfs
I don't think that's true. ext2 has certainly been around for longer,
but I'm sure ext3 wasn't generally available when I switched to
ReiserFS. Since ext3 is an easier upgrade, I would have tried it
first otherwise.
Cheers
Daniel
4:46 p.m.
I don't think that's true. ext2 has certainly been around for longer,
but I'm sure ext3 wasn't generally available when I switched to
ReiserFS. Since ext3 is an easier upgrade, I would have tried it
first otherwise.
Cheers
Daniel
You may be very right Daniel. I'm speaking form my very limited Redhat
perspective, where I think ext3 showed up long before they included
reiserfs. I'm pretty much a dummy distro user and don't have much history
outside of what RH provides as I use it in my testing of the chips we
develop.
Anecdotally I'm told that more machines run ext3 than reiserfs, but I
certainly don't know that to be true, and I have nothing against reiserfs,
since after all I'm using it on my audio drive.
My thought is that for a 2-disk audio workstation ext3 or reiserfs is
probably fine for the system drive, and I like reiserfs (or even XFS one
day...) better for my audio drive. For a single drive DAW it's a much more
personal decision I'm sure.
Cheers,
Mark
5:02 p.m.
mark,
actually, the idea is to get my jack running with --realtime
capabilities w/o a system freeze due to kernel 2.4.20 + ext3 problems.
this will hopefully fix my xrun difficulties. but what you said makes
sense, so i changed the fs of my root partition from ext3 to ext2 in the
/etc/fstab and rebooted. guess what? it still mounted as ext3! why is
that?!? just to make sure, i issued a "jackstart --realtime -d alsa -d
hw", and guess what else? the system froze, just like it should.
is there another way to mount my root partition as ext2? ...last chance
before everything gets converted! :-)
and... how would one change the journalling commit time, anyway? another
thing you mention in the 'xruns vs disk activity thread' is turning off
chrond... can you elaborate on that as well, pls?
best,
d.
Mark Knecht wrote:
If the idea here is that the xruns are caused by the
ext3 journaling
operations, wouldn't a simple test be to mount the ext3 partition as ext2
and prove it before doing all the work to move to reiserfs only to find out
it's something else? Or possibly change the journaling commit time to reduce
or increase the xruns?
5:23 p.m.
derek holzer wrote:
how would one change the journalling commit time,
anyway? another
thing you mention in the 'xruns vs disk activity thread' is turning
off chrond... can you elaborate on that as well, pls?
or how to turn off journalling entirely? i can see a direct relationship
between disk activity when i am running jack and my xruns [disk-light
blinking = xrun skipping].
d.
5:40 p.m.
On Fri, 2003-05-02 at 11:50, derek holzer wrote:
mark,
actually, the idea is to get my jack running with --realtime
capabilities w/o a system freeze due to kernel 2.4.20 + ext3 problems.
Sorry, but I missed this point as I got involved a little late and for
that I apologize. This freeze makes sense in light of the 2.4.20/ext3
issue beign known, and switchign to reiserfs would make sense too.
this will hopefully fix my xrun difficulties. but what
you said makes
sense, so i changed the fs of my root partition from ext3 to ext2 in the
/etc/fstab and rebooted. guess what? it still mounted as ext3! why is
that?!?
I do not know. It is my understand and my experience that if I change
fstab to say ext2 and reboot, as you did, then df -T will say ext2 and
not ext3. Strange.
just to make sure, i issued a "jackstart
--realtime -d alsa -d
hw", and guess what else? the system froze, just like it should.
Yep, if you are mounted ext3, then the freeze is expected.
is there another way to mount my root partition as ext2? ...last chance
before everything gets converted! :-)
Not that I know of.
and... how would one change the journalling commit
time, anyway?
Not sure. I'm looking at man pages but not spotting how to do it. I have
a vague memory of running across this before, but I don't know where.
There are some kernel patches out there dating back to last year, along
with some technical info, but possibly those require a kernel recompile
which is a pretty drastic step for you to take right now.
http://lwn.net/Articles/5584/
Here's one interesting thread I found with some interesting info:
https://listman.redhat.com/pipermail/ext3-users/2001-November/002152.html
I'll continue to poke around for what I saw. It was when I was first
researching reiserfs.
another
thing you mention in the 'xruns vs disk activity thread' is turning off
chrond... can you elaborate on that as well, pls?
ps -aux and kill the crond process as a test.
ntsysv on Redhat, turn off crond and/or anacron as a test.
I don't recommend running all the time with cron turned off, but it's
certainly OK to turn it off and see if it effects your low level xruns.
6:01 p.m.
Mark Knecht wrote:
There are some kernel patches out there dating back to
last year, along
with some technical info, but possibly those require a kernel recompile
which is a pretty drastic step for you to take right now.
http://lwn.net/Articles/5584/
Here's one interesting thread I found with some interesting info:
https://listman.redhat.com/pipermail/ext3-users/2001-November/002152.html
i tried applying the 2.4.20 + ext3 kernel patches located at
http://www.zip.com.au/~akpm/linux/ext3/
earlier this week. not only did they not apply properly, but they
screwed up my low-latency and preemptive patches too. luckily, the
errors were so bad that the kernel simply didn't compile. so i think i
will pass on any more kernel-tinkering.
so, back to plan A: with all of your helpful suggestions, i will convert
file systems this eve. the good news so far is that running PD under
Alsa as root with an -rt flag does wonders for the sound [nothing
glitches it up except moving a window on screen], so i predict much Jack
happiness with its own realtime capabilities working right.
[why not run realtime Jack as root? you may ask... on my setup, with the
new shm version of Jack, running it as root makes a logfile in the
/dev/shm directory which not only jams up the Jack/Alsa communication
somehow, but also prevents an ordinary user from starting Jack until
this logfile is removed. don't know if this is the case for any of the
rest of y'all.]
best + thx again everybody for a very informative thread,
derek
3 May
3 May
5:06 a.m.
On Sat, 3 May 2003 04:50 am, derek holzer wrote:
so i changed the fs of my root partition from ext3 to
ext2 in the
/etc/fstab and rebooted. guess what? it still mounted as ext3! why is
that?!?
The kernel needs to mount the root filesystem before it can read /etc/fstab,
which means it needs to rely on the filesystem flags to determine what type
of filesystem it is. You need to tell that partition that it doesn't have a
journal, which in effect makes it ext2. This command should do it:
tune2fs -O ^has_journal /dev/hda2
(using the correct partition, of course). However, to make things really
tricky, this can only be done on a filesystem that is umounted or read-only.
I think you mentioned you were running out of a temporary root? You might
need to use this, or boot off a rescue CD or something. On some
distributions, booting into single user mode can also lead to a ro root
filesystem, but I haven't seen that in a while.
t
--
GPG : http://n12turbo.com/tarragon/public.key
6:19 a.m.
On Fri, May 02, 2003 at 08:50:12 +0200, derek holzer wrote:
is there another way to mount my root partition as
ext2? ...last chance
before everything gets converted! :-)
Try
mount -o remount -o noload /dev/hdXX /
it wont remount as ext2 (seems like you cant easily, anoying), but I think
it might turn off the journal.
I'm sure theres a way to remount ext3 as ext2 without rebooting or
anything, but I cant remeber what it is.
- Steve
7916
days inactive
7918
days old
linux-audio-user@lists.linuxaudio.org
22 comments
9 participants
participants (9)
-
Daniel James -
derek holzer -
Eric Dantan Rzewnicki -
Frank Barknecht -
Mark Knecht -
Mark Knecht -
Paul Winkler -
Steve Harris -
Tarragon Allen