file systems

Hello a.o.l.s readers,

I'm working on a project. If it goes well, then it will result in
Slackware 10.2 being used to provide a file server to approximately 800
middle school students.

I already use Slackware 10.2 in our classroom. I work as an aide in a
special ed classroom. I use it as a file server to make life easier
communicating among our various machines. Usually, I'm doing graphics
processing to make educational materials for the students.

When I heard of a need that another school has for a file server I
volunteered my services. I know how to install, configure, boot, and
run Slackware 10.2 on a USB hard drive, so the tech-guy at the other
school provided me with a 750G hard drive in a USB enclosure.

The server that I run in the classroom is also on a hard drive in a USB
enclosure. So this is sort of a simple task. I will just recreate on
the new hard drive what I have already set up on the old one. I can
test it from any PC, then give him back the USB enclosure and drive
when I have it setup. He can keep it in the enclosre or put in in a
PC. That is not the issue.

My question, however, is this... The drive is big. I made five 100G
partitions for student data. Perhaps we will have some student data to
put in those partitions. Actually the plan is to set up 800 student
accounts, and the students will put the data there themselves.

What file systems do you bright Slackware guys and gals think I should
be using? All I have ever used is ext2. Is there something I should
know about ext3 or resierfs (spelling?) ? Would one of these be more
appropriate?

-Joe

On Thu, 21 Jun 2007 02:48:25 +0000, Joseph H. Rosevear wrote:
[color=blue]
> What file systems do you bright Slackware guys and gals think I should
> be using? All I have ever used is ext2. Is there something I should
> know about ext3 or resierfs (spelling?) ? Would one of these be more
> appropriate?[/color]

I'd go with ext3. That way you get journaling (can make data recovery
more likely in case of accident), and time-tested reliability. Some may
advise reiserfs, as it's supposedly faster when used with LOTS of small
files, but I've never seen enough difference to matter. It may not be
updated any more either due to the author's legal problems.

You can't go wrong with ext3.


--
"Ubuntu" -- an African word, meaning "Slackware is too hard for me".

Joseph H. Rosevear wrote:
[color=blue]
> What file systems do you bright Slackware guys and gals think I should
> be using? All I have ever used is ext2. Is there something I should
> know about ext3 or resierfs (spelling?) ?[/color]

If you're satisfied with ext2 on your current system, you'll find that
ext3 is essentially the same filesystem with journaling tacked on (for
improved resilience), but that you basically manage it with the same tools
as ext2 (with a small change to command-line options in most cases).

I use ext3 on my systems, and though I have none with as much space as
you're asking about, I do have some with rather large file-systems,
very long uptimes, good performance, from web servers, DNS servers,
mail servers, and netnews servers, all of which have constant demand
on the disk subsystem, though relatively low throughput demands, to an
audio workstation where I've mixed 16 tracks of recorded audio from a
1.5 hour concert, and 24 tracks of audio from an approximately 5-minute
song (and several smaller projects), both of which place more emphasis
on short-term throughput than long-term disk hammering, and neither of
which suffered at all from disk performance.

Others here have used ReiserFS with success, so I'm certainly not
agitating against its use, but I have no experience with it myself, and
so cannot comment.

--
----------------------------------------------------------------------
Sylvain Robitaille [email]syl@alcor.concordia.ca[/email]

Systems and Network analyst Concordia University
Instructional & Information Technology Montreal, Quebec, Canada
----------------------------------------------------------------------

Joe,

First of all, I wouldn't make partitions on this drive. If you install
Slackware/s root (/) on it and will use it also for data this is what
I would do;

partition 1 : / - size about 6 GB
partition 2 : for data
partition 3 : swap

The data part will be big, very big. But on top of the data partition
install lvm (logical volume manager). That way you can make as much
logical volumes as needed and can shrink or enlarge them.

As for the file system, the root partition ext3. That way you can
always boot the machine and can always access the root. The data part:
reiserfs. Because it is very stable (I use it on very big lvm - raid
arrays for several years already and never have seen any problems),
and fast.
And, as far as I know, its the only file system that allows you to
shrink or expand logical volumes.

greatings,
Charles Collette

On Thu, 21 Jun 2007 02:48:25 +0000, Joseph H. Rosevear wrote:
[color=blue]
> Hello a.o.l.s readers,
>
> I'm working on a project. If it goes well, then it will result in
> Slackware 10.2 being used to provide a file server to approximately 800
> middle school students.
>
> I already use Slackware 10.2 in our classroom. I work as an aide in a
> special ed classroom. I use it as a file server to make life easier
> communicating among our various machines. Usually, I'm doing graphics
> processing to make educational materials for the students.
>
> When I heard of a need that another school has for a file server I
> volunteered my services. I know how to install, configure, boot, and
> run Slackware 10.2 on a USB hard drive, so the tech-guy at the other
> school provided me with a 750G hard drive in a USB enclosure.
>
> The server that I run in the classroom is also on a hard drive in a USB
> enclosure. So this is sort of a simple task. I will just recreate on
> the new hard drive what I have already set up on the old one. I can
> test it from any PC, then give him back the USB enclosure and drive
> when I have it setup. He can keep it in the enclosre or put in in a
> PC. That is not the issue.
>
> My question, however, is this... The drive is big. I made five 100G
> partitions for student data. Perhaps we will have some student data to
> put in those partitions. Actually the plan is to set up 800 student
> accounts, and the students will put the data there themselves.
>
> What file systems do you bright Slackware guys and gals think I should
> be using? All I have ever used is ext2. Is there something I should
> know about ext3 or resierfs (spelling?) ? Would one of these be more
> appropriate?
>
> -Joe
>[/color]
I usually use XFS. The wikipedia has some interesting information
comparing filesystems:

[url]http://en.wikipedia.org/wiki/Comparison_of_file_systems[/url]

I agree with the conclusion that creating/destroying directories is a
relatively slow process as compared with other filesystems. Otherwise, it
works very well, especially when support for EA/ACLs is necessary
(i.e. for samba shares).

Going OT:
Portable drives are subject to being dropped, knocked off the shelf, etc.
Most of these drives have a receptable for a cable lock, and that could be
used as an anchor point to keep the drive in place.

--
Douglas Mayne

On Thu, 21 Jun 2007 08:00:40 -0600, Douglas Mayne
<doug@slack-1a.localnet> wrote:
[color=blue]
>[url]http://en.wikipedia.org/wiki/Comparison_of_file_systems[/url][/color]

Bless you! This is exactly what I've been looking for (I need a FS
where file size can exceed 4Gb in order to write DVDs).
--
buck

On Thu, 21 Jun 2007 09:11:41 -0700, buck wrote:
[color=blue]
> On Thu, 21 Jun 2007 08:00:40 -0600, Douglas Mayne
> <doug@slack-1a.localnet> wrote:
>[color=green]
>>[url]http://en.wikipedia.org/wiki/Comparison_of_file_systems[/url][/color]
>
> Bless you! This is exactly what I've been looking for (I need a FS
> where file size can exceed 4Gb in order to write DVDs).
>[/color]
I have had success writing single large files to DVDs by using k3b. Just
check the box, "udf," on the filesystem tab. But IIRC, I had trouble
exceeding 4G for any single file, but I could be wrong. If there is a 4G
limit, then device mapper's "linear" module may be a handy tool to
recombine split elements. This would avoid the need to recombine elements
before mounting the whole via loopback.

The article here:
[url]http://tinyurl.com/2bnq39[/url]
on UDF states that the overall limit to the format is 2 TB.

Note to anyone: Confirmation about the limit for individual files
would be appreciated.



Notes:
1. Use loopback file blocks to backup, using the filesystem of your choice
(internally on the loopback filesystem.)

2. If the DVD is to be bootable, make sure the boot components are
written to the disc in the first 650M, (i.e. before a large single file.)

buck wrote:
[color=blue]
> ... (I need a FS where file size can exceed 4Gb in order to write
> DVDs).[/color]

You didn't think to just try this on ext2/ext3?

--
----------------------------------------------------------------------
Sylvain Robitaille [email]syl@alcor.concordia.ca[/email]

Systems and Network analyst Concordia University
Instructional & Information Technology Montreal, Quebec, Canada
----------------------------------------------------------------------

On Thu, 21 Jun 2007 16:53:11 +0000 (UTC), Sylvain Robitaille
<syl@alcor.concordia.ca> wrote:
[color=blue]
>buck wrote:
>[color=green]
>> ... (I need a FS where file size can exceed 4Gb in order to write
>> DVDs).[/color]
>
>You didn't think to just try this on ext2/ext3?[/color]

It is mounted as a Samba share by the computer doing the DVD write, so
even though the underlying ext3 is OK, the Samba share iss not. Did
you actually READ the info at the posted URL?
--
buck

Dan C <youmustbejoking@lan.invalid> wrote:[color=blue]
> On Thu, 21 Jun 2007 02:48:25 +0000, Joseph H. Rosevear wrote:[/color]
[color=blue][color=green]
> > What file systems do you bright Slackware guys and gals think I should
> > be using? All I have ever used is ext2. Is there something I should
> > know about ext3 or resierfs (spelling?) ? Would one of these be more
> > appropriate?[/color][/color]
[color=blue]
> I'd go with ext3. That way you get journaling (can make data recovery
> more likely in case of accident), and time-tested reliability. Some may
> advise reiserfs, as it's supposedly faster when used with LOTS of small
> files, but I've never seen enough difference to matter. It may not be
> updated any more either due to the author's legal problems.[/color]
[color=blue]
> You can't go wrong with ext3.[/color]

[color=blue]
> --
> "Ubuntu" -- an African word, meaning "Slackware is too hard for me".[/color]

Sounds like ext3 is popular around here. But it is just ext2 plus
journaling, right? So I'm debating. On the one hand journaling may
help with data recovery. On the other hand, I'm familiar with ext2.

Hmmm.

-Joe

Charles Collette <collette@no-spam-xs4all.nl> wrote:[color=blue]
> Joe,[/color]
[color=blue]
> First of all, I wouldn't make partitions on this drive. If you install
> Slackware/s root (/) on it and will use it also for data this is what
> I would do;[/color]
[color=blue]
> partition 1 : / - size about 6 GB
> partition 2 : for data
> partition 3 : swap[/color]
[color=blue]
> The data part will be big, very big. But on top of the data partition
> install lvm (logical volume manager). That way you can make as much
> logical volumes as needed and can shrink or enlarge them.[/color]
[color=blue]
> As for the file system, the root partition ext3. That way you can
> always boot the machine and can always access the root. The data part:
> reiserfs. Because it is very stable (I use it on very big lvm - raid
> arrays for several years already and never have seen any problems),
> and fast.
> And, as far as I know, its the only file system that allows you to
> shrink or expand logical volumes.[/color]
[color=blue]
> greatings,
> Charles Collette[/color]

Charles,

This is a neato idea. Maybe I'll try it sometime.

-Joe

Douglas Mayne <doug@slack-1a.localnet> wrote:[color=blue]
> On Thu, 21 Jun 2007 02:48:25 +0000, Joseph H. Rosevear wrote:[/color]

[snip]
[color=blue]
> I usually use XFS. The wikipedia has some interesting information
> comparing filesystems:[/color]
[color=blue]
> [url]http://en.wikipedia.org/wiki/Comparison_of_file_systems[/url][/color]

Interesting.

[snip]
[color=blue]
> Going OT:[/color]

What is "OT"?
[color=blue]
> Portable drives are subject to being dropped, knocked off the shelf, etc.
> Most of these drives have a receptable for a cable lock, and that could be
> used as an anchor point to keep the drive in place.[/color]

Good idea. I haven't dropped one yet!
[color=blue]
> --
> Douglas Mayne[/color]

-Joe

Joseph H. Rosevear <joe@airlink9.hopto.org> wrote:

[snip]
[color=blue]
> -Joe[/color]

To all who replied,

Thanks for your interesting suggestions. I'm better educated now, but
I'm going to stick to ext2 for the time being. However, if a need
confronts us, I'm now able to suggest an alternative to ext2.

-Joe

buck wrote:
[color=blue][color=green][color=darkred]
>>> ... (I need a FS where file size can exceed 4Gb in order to write
>>> DVDs).[/color]
>>
>>You didn't think to just try this on ext2/ext3?[/color]
>
> It is mounted as a Samba share by the computer doing the DVD write, so
> even though the underlying ext3 is OK, the Samba share iss not. Did
> you actually READ the info at the posted URL?[/color]

I did. Where does it say that the file system you refer to above
("where file size can exceed 4Gb in order to write DVDs") has anything
to do with a Samba share?

--
----------------------------------------------------------------------
Sylvain Robitaille [email]syl@alcor.concordia.ca[/email]

Systems and Network analyst Concordia University
Instructional & Information Technology Montreal, Quebec, Canada
----------------------------------------------------------------------

On 2007-06-22, Joseph H. Rosevear <joe@airlink9.hopto.org> wrote:
[color=blue]
> Sounds like ext3 is popular around here. But it is just ext2 plus
> journaling, right? So I'm debating. On the one hand journaling may
> help with data recovery. On the other hand, I'm familiar with ext2.[/color]

If you are familiar with ext2, then there is little reason not to
use ext3. You can even mount an ext3 filesystem as ext2 (though of
course the journalling will not be done), and all the e2* tools
work with ext3 filesystems. There are two possible reasons to
use ext2 instead: 1) you really don't care about journalling,
say on a filesystem like /tmp, and the overhead would be wasted;
2) you're really really short on disk space, and you need the
space used by the journal for real data. If neither of these
situations apply to you, don't debate, use ext3.

--keith

--
[email]kkeller-usenet@wombat.san-francisco.ca.us[/email]
(try just my userid to email me)
AOLSFAQ=http://www.therockgarden.ca/aolsfaq.txt
see X- headers for PGP signature information

On 2007-06-22, Joseph H. Rosevear <joe@airlink9.hopto.org> wrote:
[color=blue]
> Thanks for your interesting suggestions. I'm better educated now, but
> I'm going to stick to ext2 for the time being. However, if a need
> confronts us, I'm now able to suggest an alternative to ext2.[/color]

I changed to ext3 because the occasional power blackout would too
often corrupt ext2, requiring me to fsck it and not always
successfully. Never had that problem again after changing to ext3.

nb

Douglas Mayne <doug@slack-1a.localnet> wrote:[color=blue]
> I have had success writing single large files to DVDs by using k3b. Just
> check the box, "udf," on the filesystem tab.[/color]

There is a problem with that, as the "udf" driver had problems (like
it could corrupt the whole kernel), since the 2.6.17.<something> kernel
the udf filesystem has been restricted to a maximum of 1 GB files ONLY!
As the test is not done during READing, you can still read/access larger
files, but you cannot create a file > 1 GB on an UDF fs anymore, with
those newer kernels.
--
********************************************************************
** Eef Hartman, Delft University of Technology, dept. EWI/TW **
** e-mail: [email]E.J.M.Hartman@math.tudelft.nl[/email], fax: +31-15-278 7295 **
** snail-mail: P.O. Box 5031, 2600 GA Delft, The Netherlands **
********************************************************************

On Fri, 22 Jun 2007 04:13:54 +0000 (UTC), Sylvain Robitaille
<syl@alcor.concordia.ca> wrote:
[color=blue]
>buck wrote:
>[color=green][color=darkred]
>>>> ... (I need a FS where file size can exceed 4Gb in order to write
>>>> DVDs).
>>>
>>>You didn't think to just try this on ext2/ext3?[/color]
>>
>> It is mounted as a Samba share by the computer doing the DVD write, so
>> even though the underlying ext3 is OK, the Samba share iss not. Did
>> you actually READ the info at the posted URL?[/color]
>
>I did. Where does it say that the file system you refer to above
>("where file size can exceed 4Gb in order to write DVDs") has anything
>to do with a Samba share?[/color]

First, it says ext2 & ext3 can have file sizes of 16Gb. Therefore, my
issue cannot be due my ext3 fs.

Second, Samba (although it isn't easy to find) uses FAT32 even though
it can advertize NTFS or FAT or Samba as the FS. wikipedia says FAT32
is limited to 4Gb.

So my Samba share cannot be used to write a ~4.6Gb DVD image. Now I
must decide whether or not to move the DVD writer to the Slackware
machine...

Probably because I'm too stupid, I have problems w/CIFS, so that won't
be part of my analysis.
--
buck

Joseph H. Rosevear wrote:[color=blue]
> notbob <notbob@nothome.com> wrote:[color=green]
>> On 2007-06-22, Joseph H. Rosevear <joe@airlink9.hopto.org> wrote:[/color]
>[color=green][color=darkred]
>>> Thanks for your interesting suggestions. I'm better educated now, but
>>> I'm going to stick to ext2 for the time being. However, if a need
>>> confronts us, I'm now able to suggest an alternative to ext2.[/color][/color]
>[color=green]
>> I changed to ext3 because the occasional power blackout would too
>> often corrupt ext2, requiring me to fsck it and not always
>> successfully. Never had that problem again after changing to ext3.[/color]
>
> Can I convert an ext2 system to ext3?[/color]

Yes. An ext3 filesystem is simply an ext2 filesystem with some
"invisible" files that store the journaling information.

/sbin/tune2fs -j /dev/hdx

Then change the filesystem type in /etc/fstab from ext2 to ext3. Its
recommended to unmount (and backup?) the partition before running these
commands; if you don't, a ".journal" file will be visible until the next
time e2fsck runs.
[color=blue]
> I read the section in the Filesystems-HOWTO on journaling. I didn't
> understand what it was good for. The HOWTO says that journaling file
> systems are slower than other file systems, but more reliable. The
> HOWTO explains why.
>
> So, I take back what I said. I'll give it a try.
>
> The HOWTO also said that
>
> This approach does not require a full scan of the file system,
> yielding very quick filesystem check time on large file systems,
> generally a few seconds for a multiple-gigabyte file system.
>
> Does that mean the regular check that occurs every N mounts will be
> faster? That would be cool. Especially since I have large partitions.[/color]

Many journaling filesystems do a sanity check every time a partition is
mounted, thus eliminating the need for monthly checks. That and the
improved data recovery switched me to reiserfs back in the day.

- Daniel

P.S. I *have* experienced a data corruption issue on a reiserfs
partition that was experienced several years of Slackware upgrades and
numerous power outages. By the time I noticed this, a large number of
files were unreadable; the filename was there, but the fs complained
about corruption. Possibly due to klutziness, my attempts to recover
these files resulted in over 2000 files with names like 604_641,
604_642, ... in lost+found. No idea what would have happened with
another journaling system (good idea of how this could have trashed a
non-journaling fs, though).

The point? Journaling is good, but it isn't a magic bullet.

D Herring <dherring@at.tentpost.dot.com> wrote:[color=blue][color=green]
>> Can I convert an ext2 system to ext3?[/color]
>
> Yes. An ext3 filesystem is simply an ext2 filesystem with some
> "invisible" files that store the journaling information.
>
> /sbin/tune2fs -j /dev/hdx[/color]

Preferably on a NON-mounted fs (you cannot do that for the root, of course,
except when you're booted up i.e. from CD-rom).
If the file system is mounted, the journal will be created as a hidden
file, which means that backups (and so) will see it (but you mustn't restore
it as the fs is then changed), while on a non-mounted fs the journal isn't
visible at all, just an i-node without a directory entry.

At least, that was the difference when first I changed over from ext2 to
ext3, we had some problems with those "hidden but named" journals.
[color=blue]
> if you don't, a ".journal" file will be visible until the next
> time e2fsck runs.[/color]

In our installations it STAYED visible, even after the first fsck, but
that may have changed since then.
[color=blue][color=green]
>> Does that mean the regular check that occurs every N mounts will be
>> faster? That would be cool. Especially since I have large partitions.[/color][/color]

No, the regular check is a FULL check, not using the journal, the journal
is only used on the fsck's after reboot INbetween those full checks.
You may want to change the mount count FOR those regular checks, though,
you can also do that with tune2fs (I normally put them on every 50th mount).
--
********************************************************************
** Eef Hartman, Delft University of Technology, dept. EWI/TW **
** e-mail: [email]E.J.M.Hartman@math.tudelft.nl[/email], fax: +31-15-278 7295 **
** snail-mail: P.O. Box 5031, 2600 GA Delft, The Netherlands **
********************************************************************