random experiences with kFreeBSD on Debian

[sgheeren]

Any opinions on this: http://tucobsd.blogspot.com/2010/08/apt-get-install-zfsutils.html

?

[Emmanuel Anne]

I have given it a quick try :

on my new system it panics while probing the devices at the start of the boot for the install cd !

For all the options I choose (acpi off, safe mode, everything panics after 1s of uptime only !).

The repositories are still unstable for now from what I heard, which is not a very good sign.

The idea seems good though, but it's probably too early (at least for me). You can always try to boot the install cd to see if it's working better for you !

2010/8/13 sgheeren <sghe...@hotmail.com>

Any opinions on this: http://tucobsd.blogspot.com/2010/08/apt-get-install-zfsutils.html

?

--
To post to this group, send email to zfs-...@googlegroups.com
To visit our Web site, click on http://zfs-fuse.net/

--
zfs-fuse git repository : http://rainemu.swishparty.co.uk/cgi-bin/gitweb.cgi?p=zfs;a=summary

[devsk]

Seth,

Can you elaborate a bit? What is Debian GNU/kFreeBSD? And what's the
big deal about it?

-devsk

[Gavin Chappell]

Debian GNU/kFreeBSD is a project to bring the Debian userland (including things like apt for easy software management) to a FreeBSD kernel. The big deal wrt this list is that FreeBSD can have native, in-kernel ZFS support (not via FUSE) as it's not so stringent on licensing as the Linux kernel is (I forget the specifics, but basically because FreeBSD is BSD licensed, there's nothing wrong with them including software licensed under a non-FSF compliant open source license such as Oracle's CDDL in their own kernel, whereas the GPL forbids this).

It's similar in spirit to Nexenta Core, which aimed to provide the OpenSolaris kernel with a more modern, Linux userland.

-Gavin

[sgheeren]

On 08/14/2010 09:41 AM, Gavin Chappell wrote:

Debian GNU/kFreeBSD is a project to bring the Debian userland (including things like apt for easy software management) to a FreeBSD kernel. The big deal wrt this list is that FreeBSD can have native, in-kernel ZFS support (not via FUSE) as it's not so stringent on licensing as the Linux kernel is (I forget the specifics, but basically because FreeBSD is BSD licensed, there's nothing wrong with them including software licensed under a non-FSF compliant open source license such as Oracle's CDDL in their own kernel, whereas the GPL forbids this).

It's similar in spirit to Nexenta Core, which aimed to provide the OpenSolaris kernel with a more modern, Linux userland.

-Gavin

Thx Gavin that gave me the major clue: we're just talking about FreeBSD then. Hmmm. I was hoping it was some sort of kernel virtualization layer (user-mode BSD if you will)

My gut reaction to this kind of project is usually: if I'm going to use FreeBSD, I'll just bite the bullet and grok the tools that come with the trade. But, hey, as I'm not about to switch today [1], I'll keep an eye on this one. THey might just pull an interesting stunt this time!

Seth


[1] last time I checked BSD's ZFS support lagged considerably with upstream, or zfs-fuse courtesy Emmanuels relentless updating

[Emmanuel Anne]

Finding links about it is easy. This one particularly, where debian lists the advantages of kfreebsd :

http://wiki.debian.org/Debian_GNU/kFreeBSD_why

I don't even agree with all of them :

 - devfs : except for distribution managers, it's really not a problem

 - oss : is dead, I prefer alsa, and I wouldn't want to go back !

The performance is about the same as a linux, contrary to what they thought, see the link on the page to the benchmarks.

The stability : well it panics so fast on my system that I can't really say ! But even for those systems which can actually boot it, there are areas where it's not very strong, even zfs support created panics before someone sent a set of patches...

All in all the main point of interest for me would be to test the zfs kernel implementation, but it will have to wait apparently...

2010/8/14 sgheeren <sghe...@hotmail.com>

--

[Mike Hommey]

On Sat, Aug 14, 2010 at 10:38:56AM +0200, sgheeren wrote:
> On 08/14/2010 09:41 AM, Gavin Chappell wrote:
> > Debian GNU/kFreeBSD is a project to bring the Debian userland
> > (including things like apt for easy software management) to a FreeBSD
> > kernel. The big deal wrt this list is that FreeBSD can have native,
> > in-kernel ZFS support (not via FUSE) as it's not so stringent on
> > licensing as the Linux kernel is (I forget the specifics, but
> > basically because FreeBSD is BSD licensed, there's nothing wrong with
> > them including software licensed under a non-FSF compliant open source
> > license such as Oracle's CDDL in their own kernel, whereas the GPL
> > forbids this).
> >
> > It's similar in spirit to Nexenta Core, which aimed to provide the
> > OpenSolaris kernel with a more modern, Linux userland.
> >
> > -Gavin
>
> Thx Gavin that gave me the major clue: we're just talking about FreeBSD
> then. Hmmm. I was hoping it was some sort of kernel virtualization layer
> (user-mode BSD if you will)

Debian GNU/kFreeBSD is *not* FreeBSD, and is far from it. The only
common thing really is the kernel. The whole userland is different.
FreeBSD uses a BSD userland, Debian GNU/kFreeBSD uses a GNU userland,
including glibc.

Mike

[Daniel Smedegaard Buus]

On Aug 14, 10:38 am, sgheeren <sghee...@hotmail.com> wrote:
> On 08/14/2010 09:41 AM, Gavin Chappell wrote:
>
> > Debian GNU/kFreeBSD is a project to bring the Debian userland
> > (including things like apt for easy software management) to a FreeBSD
> > kernel. The big deal wrt this list is that FreeBSD can have native,
> > in-kernel ZFS support (not via FUSE) as it's not so stringent on
> > licensing as the Linux kernel is (I forget the specifics, but
> > basically because FreeBSD is BSD licensed, there's nothing wrong with
> > them including software licensed under a non-FSF compliant open source
> > license such as Oracle's CDDL in their own kernel, whereas the GPL
> > forbids this).
>
> > It's similar in spirit to Nexenta Core, which aimed to provide the
> > OpenSolaris kernel with a more modern, Linux userland.
>
> > -Gavin
>
> Thx Gavin that gave me the major clue: we're just talking about FreeBSD
> then. Hmmm. I was hoping it was some sort of kernel virtualization layer
> (user-mode BSD if you will)
>
> My gut reaction to this kind of project is usually: if I'm going to use
> FreeBSD, I'll just bite the bullet and grok the tools that come with the
> trade. But, hey, as I'm not about to switch today [1], I'll keep an eye
> on this one. THey might just pull an interesting stunt this time!
>

I've actually been playing around with this one but never succeeded in
getting anything properly working.

The attraction to me is that in the last ~3 years I've failed to run a
Linux system on ANY computer which wouldn't grind to a virtual halt if
I had the nerve to be downloading a torrent while expanding a
large .tar.bz2. So I've been looking for alternatives, kernel-wise,
osol & bsd included (Mac OS and Win excluded though both behave
"properly" when doing disk IO), but have not succeeded in finding a
viable alternative to Linux. It's really frustrating, really, because
doing disk IO in Linuxes since like 2007 for me brings memories of old
386 machines and brewing coffee while waiting for a program to start.
I have no idea what this is related to (though I've heard about IO
sync being global regardless of the context where it's requested), I
just hate the effects :)

With osol apparently gone, it seems the viable alternatives to Linux
are getting fewer. I just hope some day disk IO won't be prioritized
higher than user IO ;)

[sgheeren]

Ok, I'm getting a taste of things by measuring peoples three-line
description here. So, obviously, I will fail to get the nuances first
time round. Thing is, as long as the kernel is BSD, at least we can
safely say that it _certainly is not linux_. That's about what I meant
to take away :)

[Emmanuel Anne]

2010/8/14 Daniel Smedegaard Buus <danie...@gmail.com>

The attraction to me is that in the last ~3 years I've failed to run a

Linux system on ANY computer which wouldn't grind to a virtual halt if
I had the nerve to be downloading a torrent while expanding a
large .tar.bz2.

Huh ? Even on my old system, I could do that without problem.

The system is slower because of the heavy disk activity but it's still very usable.

Usually when everythings "halts" this way, it's some dma problem, but nowdays it has become extremely rare.

With osol apparently gone, it seems the viable alternatives to Linux
are getting fewer. I just hope some day disk IO won't be prioritized
higher than user IO ;)

Also you can try a preempt kernel, but even a normal kernel can do that without problem. The difference is how much they slow down in this case I guess, but even the normal kernel remains very usable under heavy disk activity. 

[Daniel Smedegaard Buus]

On Aug 14, 2:47 pm, Emmanuel Anne <emmanuel.a...@gmail.com> wrote:
> 2010/8/14 Daniel Smedegaard Buus <danielb...@gmail.com>

>
> > The attraction to me is that in the last ~3 years I've failed to run a
> > Linux system on ANY computer which wouldn't grind to a virtual halt if
> > I had the nerve to be downloading a torrent while expanding a
> > large .tar.bz2.
>
> Huh ? Even on my old system, I could do that without problem.
> The system is slower because of the heavy disk activity but it's still very
> usable.
> Usually when everythings "halts" this way, it's some dma problem, but
> nowdays it has become extremely rare.
>

Which is *exactly* what was the case on MY "old system". In 2003,
doing Mandriva, I experienced that with the exact same hardware,
suddenly I could burn a DVD and do par2 creation on the same disk
simultaneously. Impossible on my old system (WinXP). I ranted about it
to my friends.

This is still the case, I can do that and still have my DVD burnt
correctly, but I can't use my puter in the meantime. Everything else
is like a DJ gone epileptic. Stutter stutter stutter hack hack hack.

> With osol apparently gone, it seems the viable alternatives to Linux
>
> > are getting fewer. I just hope some day disk IO won't be prioritized
> > higher than user IO ;)
>
> Also you can try a preempt kernel, but even a normal kernel can do that
> without problem. The difference is how much they slow down in this case I
> guess, but even the normal kernel remains very usable under heavy disk
> activity.
>

I know. And that might help. I just don't get what's happening with
the mainstream kernel. And I have no doubt that the kernel at all
times stays very happy and functional, it's just my mouse and my
firefox and my amarok (now rhythmbox after I tried switching to gnome
for the exact same reason) that is thrown into a funk.

Don't mean to start a rant thread in an entirely inappropriate place,
btw, so maybe I should just shut it :D

>
>
>

[sgheeren]

On 08/14/2010 04:58 PM, Daniel Smedegaard Buus wrote:
> Everything else
> is like a DJ gone epileptic

A poet nolens-volens :)

[Emmanuel Anne]

Yeah well can't say much about it, except that it still works fine for me, and I can burn things even easier now that they do it in "burn proof" mode (when the buffer gets empty, it just waits).

But playing music while having heavy disk activity is still ok.

I would vote for some hw problem of some kind, but I can't guess what it is exactly...

2010/8/14 sgheeren <sghe...@hotmail.com>

--
To post to this group, send email to zfs-...@googlegroups.com
To visit our Web site, click on http://zfs-fuse.net/

--

[devsk]

For folks who are troubled by large IOs (to high latency devices like
HDDs, USB devices etc.) bringing their Linux system down to its knees,
use kernel 2.6.35 with two very small patches found at

http://phoronix.com/forums/showpost.php?p=142025&postcount=38

It has fixed the issue to a large extent i.e. the system doesn't grind
to a halt anymore.

You may also wanna follow the bug: https://bugzilla.kernel.org/show_bug.cgi?id=12309

-devsk

On Aug 14, 9:11 am, Emmanuel Anne <emmanuel.a...@gmail.com> wrote:
> Yeah well can't say much about it, except that it still works fine for me,
> and I can burn things even easier now that they do it in "burn proof" mode
> (when the buffer gets empty, it just waits).
> But playing music while having heavy disk activity is still ok.
>
> I would vote for some hw problem of some kind, but I can't guess what it is
> exactly...
>

> 2010/8/14 sgheeren <sghee...@hotmail.com>

>
> > On 08/14/2010 04:58 PM, Daniel Smedegaard Buus wrote:
> > > Everything else
> > > is like a DJ gone epileptic
>
> > A poet nolens-volens :)
>
> > --
> > To post to this group, send email to zfs-...@googlegroups.com

> > To visit our Web site, click onhttp://zfs-fuse.net/

[Emmanuel Anne]

Wow impressive that some people have such problems...

I tried his test to be sure : while the dd was running in /tmp, I opened a new terminal, launched audacious2 on an mp3 file, and I am replying to this mail !

So no, definetely I don't have this problem...

Though I am using a preempt kernel, but it doesn't make a big difference normally !

(using 2.6.34.1).

2010/8/14 devsk <dev...@gmail.com>

To visit our Web site, click on http://zfs-fuse.net/

[sgheeren]

On 08/14/2010 09:10 PM, devsk wrote:
> For folks who are troubled by large IOs (to high latency devices like
> HDDs, USB devices etc.) bringing their Linux system down to its knees,
> use kernel 2.6.35 with two very small patches found at
>
> http://phoronix.com/forums/showpost.php?p=142025&postcount=38
>
> It has fixed the issue to a large extent i.e. the system doesn't grind
> to a halt anymore.
>
> You may also wanna follow the bug: https://bugzilla.kernel.org/show_bug.cgi?id=12309
>
> -devsk
>

Well.... that came out of nowhere ?! Splendid info, thanks, but it might
have been appropriate to post a new thread for this. Well, look at me
nagging :) I'm adding this info to issue #82 where I'm kind-of giving
Bryan a hard time convincing me of performance problems... Your
description looks surprisingly similar, so let's see what he can do with
the info :)

Cheers,
Seth

[devsk]

Well, that problem is 4 years old! The first bug 7372 got too big and
a mess, so a new one was created. That one's now got 500+ comments on
it. So, you bet there are real people hitting that issue!

Preempt, HZ=1000, CFQ, deadline, various variety of swappiness,
vfs_cache_pressure, dirty_ratio etc. have all been tried over the
course of last 4 years. Nothing helped. And then, this guy from Intel
comes along and creates a small patch. And things are looking better!

-devsk

[Daniel Smedegaard Buus]

On Aug 14, 10:36 pm, devsk <dev...@gmail.com> wrote:
> Well, that problem is 4 years old! The first bug 7372 got too big and
> a mess, so a new one was created. That one's now got 500+ comments on
> it. So, you bet there are real people hitting that issue!
>
> Preempt, HZ=1000, CFQ, deadline, various variety of swappiness,
> vfs_cache_pressure, dirty_ratio etc. have all been tried over the
> course of last 4 years. Nothing helped. And then, this guy from Intel
> comes along and creates a small patch. And things are looking better!
>
> -devsk
>

Oh my GOD, I can't believe this is getting fixed!!! I was starting to
get seriously depressed about this. I actually looked towards Windows
7 at work at one time as it had gotten so bad that each new kernel
update (I was running Lucid Alpha, hoping the performance there was
better). Of course that didn't pan out ;)

This has been YEARS of me being frustrated, just thinking "why the
hell are we seeing new version of KDE when this bug is so severe and
seemingly ignored?".

I have tried SO many distros, even more exotic ones like Sabayon and
PC-BSD, to somehow get to that UI experience I had years ago with
Linux. It's been like a lover getting fat. You stick with her because
she still has so many other qualities, and you just wouldn't be able
to live without her, but damn it, she's been hard dragging around.

I even switched to Gnome recently, thinking what the hell, it might be
related to KDE or even KDM, and actually it seems a little better,
though not much. Maybe its default scheduler priorities are different,
I don't know.

Don't care, cause we're getting a fix?!?!?! I can't believe it! :D Me
so happy!

Any idea when we'll see this in Ubuntu-based distros?

Cheers all, haaaapppyy day indeed!!! :D

[sgheeren]

On 08/15/2010 09:50 AM, Daniel Smedegaard Buus wrote:
> I even switched to Gnome recently, thinking what the hell, it might be
> related to KDE or even KDM, and actually it seems a little better,
> though not much. Maybe its default scheduler priorities are different,
> I don't know.
>

Traditionally, Gnome has been hitting the disks somewhat less hard (or
rather: KDE has a track record of not optimizing for disk access)
The gap is narrowing, these days

For truly lightweight see a minimal desktop like Xfce.

> Any idea when we'll see this in Ubuntu-based distros?
>

You can monitor Ubuntu's kernel repo, and smidge your own package.

git://kernel.ubuntu.com/ubuntu/ubuntu-lucid.git

I'm sure there is a place where you could probably vote this patch up a
bit... I wouldn't know where

Seth

[Daniel Smedegaard Buus]

On Aug 17, 10:18 am, sgheeren <sghee...@hotmail.com> wrote:
> On 08/15/2010 09:50 AM, Daniel Smedegaard Buus wrote:
>
> > I even switched to Gnome recently, thinking what the hell, it might be
> > related to KDE or even KDM, and actually it seems a little better,
> > though not much. Maybe its default scheduler priorities are different,
> > I don't know.
>
> Traditionally, Gnome has been hitting the disks somewhat less hard (or
> rather: KDE has a track record of not optimizing for disk access)
> The gap is narrowing, these days
>

That was pretty much my feeling after working with it for awhile. It
seem to rarely hits as violent IO thrash extremes (thrash extremes!
YIAH! :D ) as with KDE. However, as a number of other people have also
pointed out, the problem seems to increase over time (reading up on
it, I guess because there are simple more pages to evict when you try
to do something new).

> For truly lightweight see a minimal desktop like Xfce.
>
> > Any idea when we'll see this in Ubuntu-based distros?
>
> You can monitor Ubuntu's kernel repo, and smidge your own package.
>
>         git://kernel.ubuntu.com/ubuntu/ubuntu-lucid.git
>
> I'm sure there is a place where you could probably vote this patch up a
> bit... I wouldn't know where
>

Thanks for the link :) I actually went ahead and ran for the zen
kernel git repo. I had an initial attempt here on my home puter (the
one with the ZFS titanic pool), but I guess it's been too long since I
last compiled my own kernel (Mandriva, mid-2000s), so I kinda fudged
it up around the make install part... Also, wow! The compile time has
increased since last :O

Anyway, I read up on it, what with the make-kpkg thing and all that,
and first got me a zen kernel on my laptop (which I haven't stress
tested yet, but the new kernel brings lots of improvements with
regards to all the nice buttons (ACPI?), so I can now switch on and
off my wireless and BT, and all that).

I then proceeded to do the same on my work computer. Compiling the
thing, I thought my quadcore might just throw in the towel, it was
that unresponsive (I also thought, "How can such a bad 'bug' be
overlooked by kernel developers for so long if this is what happens
when you compile a kernel?"). Until I did a swapoff -a, actually. It
immediately started being responsive again (_while_ swapping
everything back into main memory mind you), I guess because no more
pages were being evicted to swap.

On reboot, I did a triple simultaneous check on major torrent
downloads, unpacked a ½G lzma source archive, started burning a DVD at
10MB/s and fired up Firefox. Everything was smooth sailing. Compiz was
ninja-styling it big time, everything I clicked just went, "yiah,
what's up, let's go", not a single stutter of the mouse or interface.
It was like being dragged out of four years ago, only with a ten times
as powerful box. My colleague thought I was going insane, I was that
jubilant :)

Of course there are other "tweaks" in the zen kernel, but no way that
could've brought this kind of change. This patch just saved Linux. I'm
serious :) I could care less for everything else now, retarded Redmond
can stuff their "sorry-you-can't-have-raw-access-to-your-disks-because-
someone-wrote-some-malware" Windows 7 up their bongholes, and Oracle
can keep their Solaris proprietary and kill the community, cause Tux
is back, baby! And this time he doesn't even know what it is, cause
he's so far ahead he forgot about the competition!

Yiah!!! :D hehehe

> Seth

[sgheeren]

On 08/17/2010 05:20 PM, Daniel Smedegaard Buus wrote:
> Until I did a swapoff -a, actually. It
> immediately started being responsive again (_while_ swapping
> everything back into main memory mind you), I guess because no more
> pages were being evicted to swap.
>

Oh. I stopped using swap in 2007 I think. No. Swap. Whatsoever. Ever.

Only swap is on a DLINK NAS appliance. It runs linux in 32 Mb :)

[sgheeren]

On 08/17/2010 05:20 PM, Daniel Smedegaard Buus wrote:
>
> Thanks for the link :) I actually went ahead and ran for the zen
> kernel git repo.

> I had an initial attempt here on my home puter [...]

> what with the make-kpkg thing and all that,
>

> I then proceeded to do the same on my work computer. Compiling the
> thing,
>
>

Why? make-kpkg should give you a deb package. Simple matter of 'dpkg -i
linux......deb' on your other puter methinks? Or are they different
platforms?

> I was that
> jubilant :)
>

Good to hear. Stay jubilant and have a beer!
Seth

[Daniel Smedegaard Buus]

LOL :D That is awesome!

And that's effectively put you in the safe spot for the last Simpson-
esque handful of years, too, so you haven't even experienced my woes!

Damn you ;)

[Daniel Smedegaard Buus]

On Aug 17, 5:37 pm, sgheeren <sghee...@hotmail.com> wrote:
> On 08/17/2010 05:20 PM, Daniel Smedegaard Buus wrote:
>
>
>
> > Thanks for the link :) I actually went ahead and ran for the zen
> > kernel git repo.
> > I had an initial attempt here on my home puter [...]
> > what with the make-kpkg thing and all that,
>
> > I then proceeded to do the same on my work computer. Compiling the
> > thing,
>
> Why? make-kpkg should give you a deb package. Simple matter of 'dpkg -i
> linux......deb' on your other puter methinks? Or are they different
> platforms?

Oh yes, but I had to learn that ;)

>  I was that
> > jubilant :)
>
> Good to hear. Stay jubilant and have a beer!

No more beers, drank them all this weekend :)

> Seth

[Daniel Smedegaard Buus]

On Aug 17, 5:41 pm, Daniel Smedegaard Buus <danielb...@gmail.com>
wrote:

> On Aug 17, 5:37 pm, sgheeren <sghee...@hotmail.com> wrote:
>
> > On 08/17/2010 05:20 PM, Daniel Smedegaard Buus wrote:
>
> > > Thanks for the link :) I actually went ahead and ran for the zen
> > > kernel git repo.
> > > I had an initial attempt here on my home puter [...]
> > > what with the make-kpkg thing and all that,
>
> > > I then proceeded to do the same on my work computer. Compiling the
> > > thing,
>
> > Why? make-kpkg should give you a deb package. Simple matter of 'dpkg -i
> > linux......deb' on your other puter methinks? Or are they different
> > platforms?
>

OH! You mean because I had already built it?

Well, yeah, that's what I thought, actually. I scp'ed the built ones
from my Kubuntu box at work to my Mint box at home, but when dpkg'ing
them it gave some weird errors about some symlinks pointing into the
abyss... So I just did a rebuild, and no more errors.

Did throw me off, actually. I mean, why make a .deb that isn't
portable (within reason, of course, architectures and such).

Maybe defaults are to make one that exactly matches stuff like SSE3 or
whatever, and my home puter doesn't have that? I mean, it does have
one of those cheap dualcores without VT-x, so it might have other
shortcomings, too :)

[devsk]

Seth,

One of the tricks to avoid the disk thrashing by KDE is to set

KDETMP=/var/tmp
KDEVARTMP=/var/tmp (or to whatever tmpfs you already have)

and then mount the /var/tmp in tmpfs. The only thing you lose is the
ability to carry old cache (doesn't impact your ability to save
settings) from one KDE session to next when a reboot occurs. Small
price to pay for avoiding IO!

-devsk

[Emmanuel Anne]

Good, but which size to reserve in tmpfs for that ?

I guess you'd better not run out of space for this one... 100 Mb might be enough though ?

2010/8/18 devsk <dev...@gmail.com>

--
To post to this group, send email to zfs-...@googlegroups.com

To visit our Web site, click on http://zfs-fuse.net/

[sgheeren]

On 08/18/2010 08:02 AM, devsk wrote:
> Seth,
>
> One of the tricks to avoid the disk thrashing by KDE is to set
>
> KDETMP=/var/tmp
> KDEVARTMP=/var/tmp (or to whatever tmpfs you already have)
>
> and then mount the /var/tmp in tmpfs. The only thing you lose is the
> ability to carry old cache (doesn't impact your ability to save
> settings) from one KDE session to next when a reboot occurs. Small
> price to pay for avoiding IO!
>
> -devsk
>

Splendid tip. Unfortunately I stopped using KDE before (once they pushed
the KDE 4 experience on me way before it became usable/pretty).
I'm using the same type of tweaks with Opera/Firefox. Also enhances the
privacy and especially the wear on my SSD drives.

Needless to say, with two SSDs in RAID0, 8 Gb RAM and a Q9550 things are
_really_ flying on my workstation. I'm so spoiled!

[sgheeren]

On 08/18/2010 09:26 AM, Emmanuel Anne wrote:

Good, but which size to reserve in tmpfs for that ?

I guess you'd better not run out of space for this one... 100 Mb might be enough though ?

Good thing about tmpfs is no need to reserve up front. Only when unrarring DVD iso's [1] I ever issue a

      sudo mount -o remount,size=6G /tmp

to temporarily boost the capacity. By default my system (8Gb) sets the limit at 4Gb. Note that no space is allocated unless files are actually there.

I always run my test pool configurations directly on tmpfs. I can have multiple multi-Terabyte pools all inside tmpfs and fill them with max. 4Gb of data without running into any limits :)


[1] for backup purposes

[Emmanuel Anne]

Ok, thanks, I should try that then !

2010/8/18 sgheeren <sghe...@hotmail.com>

[devsk]

Seth,

You should try KDE 4.5. Its silky smooth...:-D

No, really! 4.5 is definitely closest to 3.5.10 in functionality yet
and very stable!

Emmanuel: I have never seen it use more than 200MB, even with 50000
command history in konsole. Don't make your konsole history
"Unlimited" though if you go tmpfs way. Bad things will happen when
kdecache fills up...:-)

But as Seth said, the default of 50% of RAM is good, unless you have
128MB of RAM....:-)

-devsk