FAQ Search Today's Posts Mark Forums Read
» Video Reviews

» Linux Archive

Linux-archive is a website aiming to archive linux email lists and to make them easily accessible for linux users/developers.


» Sponsor

» Partners

» Sponsor

Go Back   Linux Archive > Redhat > Fedora User

 
 
LinkBack Thread Tools
 
Old 11-06-2011, 01:20 PM
Fernando Cassia
 
Default Is btrfs ready to be default fs in F17 ?

On Sun, Nov 6, 2011 at 07:30, Alan Cox <alan@lxorguk.ukuu.org.uk> wrote:

Plus in a funny way btrfs is now in part un-needed,
Funny how so many people think the btrfs designers are morons.
Or as if it were a half-baked project cooked hours ago.BTRFS is five years old.

"In 2008, the principal developer of the*ext3*and*ext4*file systems,*Theodore Ts'o, stated that although*ext4*has improved features, it is not a major advance, it uses old technology, and is a stop-gap; Ts'o believes that Btrfs is the better direction because "it offers improvements in scalability, reliability, and ease of management".[4]*Btrfs also has "a number of the same design ideas thatreiser3/4*had".[5]"


http://en.wikipedia.org/wiki/Btrfs

BTRfsck is coming along nicely
http://www.mail-archive.com/linux-btrfs@vger.kernel.org/msg11836.html

so, let´s cut the FUD, please...
FC
--
"The purpose of computing is insight, not numbers."
Richard Hamming - http://en.wikipedia.org/wiki/Hamming_code




--
users mailing list
users@lists.fedoraproject.org
To unsubscribe or change subscription options:
https://admin.fedoraproject.org/mailman/listinfo/users
Guidelines: http://fedoraproject.org/wiki/Mailing_list_guidelines
 
Old 11-06-2011, 02:40 PM
Alan Cox
 
Default Is btrfs ready to be default fs in F17 ?

On Sun, 6 Nov 2011 11:20:22 -0300
Fernando Cassia <fcassia@gmail.com> wrote:

> On Sun, Nov 6, 2011 at 07:30, Alan Cox <alan@lxorguk.ukuu.org.uk> wrote:
>
> > Plus in a funny way btrfs is now in part un-needed,
>
>
> Funny how so many people think the btrfs designers are morons.

Funny how some people read strange things into discussions for their own
goals ...

> "In 2008, the principal developer of the ext3<http://en.wikipedia.org/wiki/Ext3>

In 2008 SSDs were basically laboratory dreams or infeasibly expensive toys

> management".[4]<http://en.wikipedia.org/wiki/Btrfs#cite_note-3> Btrfs
> also has "a number of the same design ideas
> thatreiser3<http://en.wikipedia.org/wiki/ReiserFS>

I'd point out two things

1. Wackypedia is not a reference source of any grade, its a sewer of
consensus opinion

2. Reiserfs also had serious problems with fsck handling and stability in
the face of failure cases. It also had a nasty tendancy to degrade in
performance over time. SuSE moved from Reiserfs to ext* having embraced
it as the big new technology of the future and discovering it wasn't.

The reiserfs world really indicates some of the problems. B tree based
file systems ought to be wonderful things. They can do a lot of stuff a
traditional cylinder group based file system cannot do nicely. But
they've also proved to be very fragile, very hard to get right and very
difficult to performance tune for long term stability.

It *can* be done - NetApp have proved that and have done this for years.
Their environment is quite special in some ways but it's a general
purpose fs with snapshots and the like that works.

> BTRfsck is coming along nicely
> http://www.mail-archive.com/linux-btrfs@vger.kernel.org/msg11836.html

It is not production ready. There was a talk covering btrfs fsck stuff in
Prague last week.

> so, let´s cut the FUD, please...

The only FUD source appears to be you right now.

At this point in time btrfs tends to break on sudden power failure cases
or media errors. It has an fsck that just about made 'can do a demo'
status two weeks ago. These are not good attributes for your default file
system.

Yes it'll change eventually perhaps - Oracle, and Red Hat and others
according to Chris Mason are committed to making it work. However using
Fedora users valuable data as cannon fodder is not how it should happen.

I would suggest http://lwn.net/Articles/462543/ is better reading matter
than Wonkypedia.

Even then I suspect for most users ext3/4 is going to continue to be
significantly more robust, easier to fix when it fails, and has a longer
track history under load. I value my data and the ability to recover it
far more than funky snapshotting features.

Alan
--
users mailing list
users@lists.fedoraproject.org
To unsubscribe or change subscription options:
https://admin.fedoraproject.org/mailman/listinfo/users
Guidelines: http://fedoraproject.org/wiki/Mailing_list_guidelines
 
Old 11-06-2011, 02:51 PM
Fernando Cassia
 
Default Is btrfs ready to be default fs in F17 ?

On Sun, Nov 6, 2011 at 12:40, Alan Cox <alan@lxorguk.ukuu.org.uk> wrote:

B tree based

file systems ought to be wonderful things. They can do a lot of stuff a

traditional cylinder group based file system cannot do nicely. But

they've also proved to be very fragile, very hard to get right and very

difficult to performance tune for long term stability.

I´ve run Btree based HPFS for over a decade...
http://pages.cs.wisc.edu/~bolo/shipyard/hpfs.html

FC
--
"The purpose of computing is insight, not numbers."
Richard Hamming - http://en.wikipedia.org/wiki/Hamming_code




--
users mailing list
users@lists.fedoraproject.org
To unsubscribe or change subscription options:
https://admin.fedoraproject.org/mailman/listinfo/users
Guidelines: http://fedoraproject.org/wiki/Mailing_list_guidelines
 
Old 11-06-2011, 02:52 PM
Fernando Cassia
 
Default Is btrfs ready to be default fs in F17 ?

On Sun, Nov 6, 2011 at 12:40, Alan Cox <alan@lxorguk.ukuu.org.uk> wrote:

Yes it'll change eventually perhaps - Oracle, and Red Hat and others

according to Chris Mason are committed to making it work.
That´s what I care about. You seem more focused on scaring people with the fear that it might not work, eat users´ data and their children, too. "perhaps".

Well, perhaps yes, perhaps not. You´re making assumptions about the quality of code yet to come at some point in the future.
FC
--
"The purpose of computing is insight, not numbers."

Richard Hamming - http://en.wikipedia.org/wiki/Hamming_code



--
users mailing list
users@lists.fedoraproject.org
To unsubscribe or change subscription options:
https://admin.fedoraproject.org/mailman/listinfo/users
Guidelines: http://fedoraproject.org/wiki/Mailing_list_guidelines
 
Old 11-06-2011, 04:11 PM
Alan Cox
 
Default Is btrfs ready to be default fs in F17 ?

> That´s what I care about. You seem more focused on scaring people with the
> fear that it might not work, eat users´ data and their children, too.
> "perhaps".

Right now btrfs is not a production fs. In 18 months maybe, in six months
the bits like fsck may exist in a usable form but I don't believe there
will be enough testing history to be sure.

It took a very very long time to get the reiserfs fsck usable and it was
never really 100%. Btrfs is a similar challenge, if not in fact a larger
one in some respects.

> Well, perhaps yes, perhaps not. You´re making assumptions about the quality
> of code yet to come at some point in the future.

Based on the history of that code, the history of such file systems and
what we know from extrapolating. "Perhaps yes, perhaps not" is not how I
want to think about whether my data will be recoverable.

The underlying problems are simple - btree file systems are very hard to
get robust, btree file system fsck tools are very complicated beasts.

It's not about code quality, any more than 'not using prototype
aeroplanes for revenue service' is about design quality. It's about
complex systems and failure patterns combined with the failure cases
having very bad effects.

Alan
--
users mailing list
users@lists.fedoraproject.org
To unsubscribe or change subscription options:
https://admin.fedoraproject.org/mailman/listinfo/users
Guidelines: http://fedoraproject.org/wiki/Mailing_list_guidelines
 
Old 11-06-2011, 04:18 PM
Genes MailLists
 
Default Is btrfs ready to be default fs in F17 ?

On 11/06/2011 12:11 PM, Alan Cox wrote:
>
> Right now btrfs is not a production fs. In 18 months maybe, in six months
> the bits like fsck may exist in a usable form but I don't believe there
> will be enough testing history to be sure.
>
> It took a very very long time to get the reiserfs fsck usable and it was
> never really 100%. Btrfs is a similar challenge, if not in fact a larger
> one in some respects.
>
>

Alan is dead on - btfrs should absolutely not be the default fs in
fedora / RHEL - its not ready yet, and Chris Mason seems to say exactly
the same thing - and he should know.

That said - of course anyone willing to use it - great - go ahead the
more testers the better :-)

gene


--
users mailing list
users@lists.fedoraproject.org
To unsubscribe or change subscription options:
https://admin.fedoraproject.org/mailman/listinfo/users
Guidelines: http://fedoraproject.org/wiki/Mailing_list_guidelines
 
Old 11-06-2011, 04:33 PM
Alan Cox
 
Default Is btrfs ready to be default fs in F17 ?

> I´ve run Btree based HPFS for over a decade...
> http://pages.cs.wisc.edu/~bolo/shipyard/hpfs.html

HPFS is for the most part a traditional Unix file system with cylinder
groups.

You have superblock (block 16), and then 8MB bands of data with
allocation bitmaps between them. Sound familiar ? It also does
pre-allocation and give back (as ext3/4 now do with reservations) but has
no journalling.

So its really a BSD FFS like file system with fancier metadata. That's
completely different to a b-tree/b+tree file system like reiserfs3 or
btrfs. In particular it uses btrees within objects only. So corrupting
the btree of an FNODE loses you a file and your coherency checking
problem for the tree is internal to each object. A DIRBLK is perhaps
slightly worse news but you've still got the FNODE so you can find the
contents of everything.

Apple HFS/HFS+ has a similar sort of design. Ask any long term MacOS
user about the dreaded "Invalid B-tree node" message which is usually
followed by "Error: File system verify or repair failed"

Alan
--
users mailing list
users@lists.fedoraproject.org
To unsubscribe or change subscription options:
https://admin.fedoraproject.org/mailman/listinfo/users
Guidelines: http://fedoraproject.org/wiki/Mailing_list_guidelines
 
Old 11-06-2011, 07:27 PM
Heinz Diehl
 
Default Is btrfs ready to be default fs in F17 ?

On 06.11.2011, Alan Cox wrote:

> B tree based file systems ought to be wonderful things. They can do
> a lot of stuff a traditional cylinder group based file system cannot
> do nicely. But they've also proved to be very fragile, very hard
> to get right and very difficult to performance tune for long term
> stability.

> It *can* be done - NetApp have proved that and have done this for years.

And there's also XFS... :-)

--
users mailing list
users@lists.fedoraproject.org
To unsubscribe or change subscription options:
https://admin.fedoraproject.org/mailman/listinfo/users
Guidelines: http://fedoraproject.org/wiki/Mailing_list_guidelines
 
Old 11-07-2011, 11:36 AM
Rudolf Kastl
 
Default Is btrfs ready to be default fs in F17 ?

2011/11/6 Fernando Cassia <fcassia@gmail.com>:
>
>
> On Sun, Nov 6, 2011 at 12:40, Alan Cox <alan@lxorguk.ukuu.org.uk> wrote:
>>
>> Yes it'll change eventually perhaps - Oracle, and Red Hat and others
>> according to Chris Mason are committed to making it work.
>
> That´s what I care about. You seem more focused on scaring people with the
> fear that it might not work, eat users´ data and their children, too.
> "perhaps".
> Well, perhaps yes, perhaps not. You´re making assumptions about the quality
> of code yet to come at some point in the future.
> FC

For a critical component like a filesystem, it should be considered
not working until really tested and proven otherwise for anything not
beeing a test machine that is ready for beeing wiped. You might not
care for your personal box... others do... others also run "production
machines". It is not about beeing scared but having realistic
expectations of code that hasnt seen field testing at all.

kind regards,
Rudolf Kastl

> --
> "The purpose of computing is insight, not numbers."
> Richard Hamming - http://en.wikipedia.org/wiki/Hamming_code
>
>
> --
> users mailing list
> users@lists.fedoraproject.org
> To unsubscribe or change subscription options:
> https://admin.fedoraproject.org/mailman/listinfo/users
> Guidelines: http://fedoraproject.org/wiki/Mailing_list_guidelines
>
>
--
users mailing list
users@lists.fedoraproject.org
To unsubscribe or change subscription options:
https://admin.fedoraproject.org/mailman/listinfo/users
Guidelines: http://fedoraproject.org/wiki/Mailing_list_guidelines
 
Old 11-07-2011, 01:38 PM
Fernando Cassia
 
Default Is btrfs ready to be default fs in F17 ?

On Mon, Nov 7, 2011 at 09:36, Rudolf Kastl <che666@gmail.com> wrote:



For a critical component like a filesystem, it should be considered

not working until really tested and proven otherwise for anything not

beeing a test machine that is ready for beeing wiped.
Oh sorry, I thought Fedora was all about running bleeding edge stuff...
FC--
"The purpose of computing is insight, not numbers."

Richard Hamming - http://en.wikipedia.org/wiki/Hamming_code


--
users mailing list
users@lists.fedoraproject.org
To unsubscribe or change subscription options:
https://admin.fedoraproject.org/mailman/listinfo/users
Guidelines: http://fedoraproject.org/wiki/Mailing_list_guidelines
 

Thread Tools




All times are GMT. The time now is 03:42 AM.

VBulletin, Copyright ©2000 - 2014, Jelsoft Enterprises Ltd.
Content Relevant URLs by vBSEO ©2007, Crawlability, Inc.
Copyright ©2007 - 2008, www.linux-archive.org