BTRFS 0.17 in rawhide (at some point)

[Josef Bacik]

Hello,

I've just updated btrfs-progs to 0.17 in rawhide, and will be doing
the same for F-10 and F-9 shortly.  If you are using the rawhide
kernel then you will be able to test with the newly merged btrfs
filesystem.  All testing is greatly appreciated.  Here is a wiki page
which outlines some of the basic commands you can use with btrfs

http://btrfs.wiki.kernel.org/index.php/Getting_started

Btrfs has built in snapshotting, subvolumes and RAID support, so lots
of fun new things to tinker with.  A few disclaimers

1) Goes without saying, but this is a brand new filesystem that is not
finished yet, so treat it as if it would rather eat your hand than
keep your data
2) SELinux support is not quite there yet.  It does have xattr
support, and there is a patch available for support, but its not
upstream yet, so not in rawhide yet.
3) The disk format is not _supposed_ to change, but if there is some
major flaw found then it will change, so plan for the worst.
4) You cannot fill the disk about 85%.  This is not a bug, its a hard
limit, and on the TODO list to fix, do not be alarmed if it cuts out.

Please feel free to file a bugzilla if something goes horribly wrong,
stability and finishing out the remaining features is what our main
focus is now.  Thanks,

Josef

[Rahul Sundaram]

Josef Bacik wrote:
>  Hello,
>  
>  I've just updated btrfs-progs to 0.17 in rawhide, and will be doing
>  the same for F-10 and F-9 shortly.  If you are using the rawhide
>  kernel then you will be able to test with the newly merged btrfs
>  filesystem.  All testing is greatly appreciated.  Here is a wiki page
>  which outlines some of the basic commands you can use with btrfs
>  
>  http://btrfs.wiki.kernel.org/index.php/Getting_started
>  
>  Btrfs has built in snapshotting, subvolumes and RAID support, so lots
>  of fun new things to tinker with.   
I know this is very early but just to get an idea of the status, has 
anyone looked into:

  * Anaconda support for Fedora 11
  * Simple home user backup that takes advantage of snapshots
  * RPM/Yum integration for rollbacks

Rahul

[Josh Boyer]

On Tue, Jan 13, 2009 at 12:53:06PM +0530, Rahul Sundaram wrote:
>  Josef Bacik wrote:
> > Hello,
> >
> > I've just updated btrfs-progs to 0.17 in rawhide, and will be doing
> > the same for F-10 and F-9 shortly.  If you are using the rawhide
> > kernel then you will be able to test with the newly merged btrfs
> > filesystem.  All testing is greatly appreciated.  Here is a wiki page
> > which outlines some of the basic commands you can use with btrfs
> >
> > http://btrfs.wiki.kernel.org/index.php/Getting_started
> >
> > Btrfs has built in snapshotting, subvolumes and RAID support, so lots
> > of fun new things to tinker with.  
>
>  I know this is very early but just to get an idea of the status, has  
>  anyone looked into:
>
>   * Anaconda support for Fedora 11 
There was an Anaconda session at FUDCon.  This wasn't mentioned at all.
I think there are enough things to do for F11 at the moment, and given
that btrfs isn't even in a stable release yet it's probably premature
to talk about supporting it for installs.

>   * Simple home user backup that takes advantage of snapshots
>   * RPM/Yum integration for rollbacks 
All of these seem like F12 topics to me.

josh

[Josef Bacik]

On Tue, Jan 13, 2009 at 7:31 AM, Josh Boyer <jwboyer(a)gmail.com&gt; wrote:
>  On Tue, Jan 13, 2009 at 12:53:06PM +0530, Rahul Sundaram wrote:
> > Josef Bacik wrote:
> >> Hello,
> >>
> >> I've just updated btrfs-progs to 0.17 in rawhide, and will be doing
> >> the same for F-10 and F-9 shortly.  If you are using the rawhide
> >> kernel then you will be able to test with the newly merged btrfs
> >> filesystem.  All testing is greatly appreciated.  Here is a wiki page
> >> which outlines some of the basic commands you can use with btrfs
> >>
> >> http://btrfs.wiki.kernel.org/index.php/Getting_started
> >>
> >> Btrfs has built in snapshotting, subvolumes and RAID support, so lots
> >> of fun new things to tinker with.
> >
> > I know this is very early but just to get an idea of the status, has
> > anyone looked into:
> >
> >  * Anaconda support for Fedora 11
>
>  There was an Anaconda session at FUDCon.  This wasn't mentioned at all.
>  I think there are enough things to do for F11 at the moment, and given
>  that btrfs isn't even in a stable release yet it's probably premature
>  to talk about supporting it for installs.
>  
Eric has sent along an anaconda patch.  The thinking is that there
will be some magic incantation to turn it on for F11, kind of like
linux iamanext4developer when ext4 wasn't quite ready.  Thanks,

Josef

[Thorsten Leemhuis]

On 13.01.2009 13:31, Josh Boyer wrote:
>  On Tue, Jan 13, 2009 at 12:53:06PM +0530, Rahul Sundaram wrote:
> > Josef Bacik wrote:
> >> Btrfs has built in snapshotting, subvolumes and RAID support, so lots
> >> of fun new things to tinker with.  
> > I know this is very early but just to get an idea of the status, has  
> > anyone looked into:
> >  * Anaconda support for Fedora 11
>  There was an Anaconda session at FUDCon.  This wasn't mentioned at all.
>  I think there are enough things to do for F11 at the moment, 
FYI, patches from Sandeen seems to be in the works:

https://www.redhat.com/archives/anaconda-devel-list/2009-January/msg00077...
https://www.redhat.com/archives/anaconda-devel-list/2009-January/msg00091...

>  [...] 
CU
knurd

[Tomasz Torcz]

On Tue, Jan 13, 2009 at 07:31:59AM -0500, Josh Boyer wrote:
>  >> http://btrfs.wiki.kernel.org/index.php/Getting_started
>  >>
>  >> Btrfs has built in snapshotting, subvolumes and RAID support, so lots
>  >> of fun new things to tinker with.   
>  >  * Simple home user backup that takes advantage of snapshots
>  >  * RPM/Yum integration for rollbacks
>  
>  All of these seem like F12 topics to me. 
  As for backup, the code is already written (cron job snapshotting and
nautilus extension for GUI). It's here:
http://src.opensolaris.org/source/xref/jds/time-slider/

  It may need adjustment to uset btrfs snapshotting mechanism, but
apart from this logic and UI is implemented.


-- 
Tomasz Torcz                                                       72->|   80->|
xmpp: zdzichubg(a)chrome.pl                                          72->|   80->|

[Seth Vidal]

On Tue, 2009-01-13 at 12:53 +0530, Rahul Sundaram wrote:
>  Josef Bacik wrote:
>  > Hello,
>  > 
>  > I've just updated btrfs-progs to 0.17 in rawhide, and will be doing
>  > the same for F-10 and F-9 shortly.  If you are using the rawhide
>  > kernel then you will be able to test with the newly merged btrfs
>  > filesystem.  All testing is greatly appreciated.  Here is a wiki page
>  > which outlines some of the basic commands you can use with btrfs
>  > 
>  > http://btrfs.wiki.kernel.org/index.php/Getting_started
>  > 
>  > Btrfs has built in snapshotting, subvolumes and RAID support, so lots
>  > of fun new things to tinker with.  
>  
>  I know this is very early but just to get an idea of the status, has 
>  anyone looked into:
>  
>    * RPM/Yum integration for rollbacks 
the last one would be implemented in exactly the same way lvm snapshots
would - via a yum plugin running before a transaction.

-sv

[David Woodhouse]

On Mon, 2009-01-12 at 16:16 -0500, Josef Bacik wrote:
>  Hello,
>  
>  I've just updated btrfs-progs to 0.17 in rawhide, and will be doing
>  the same for F-10 and F-9 shortly.  
Can we have nfs-utils updated too, please? The one we have in F-9
predates my patches to use statfs() to obtain a fsid for exporting the
file system. Not sure about F-10, but 1.1.4 in rawhide is OK.

The failure mode is that mountd returns a root fh which the client then
immediately tries to use, the kernel passes it up to mountd to be
resolved... and mountd disavows all knowledge of it. Stupidly.

This results in either the client's mount saying 'ESTALE', or just going
into a loop trying over and over again, while the server repeatedly logs
that it has authenticated a mount attempt...

The workaround for that is to use an explicit fsid= option
in /etc/exports for btrfs file systems, but updating nfs-utils would be
better.

-- 
David Woodhouse                            Open Source Technology Centre
David.Woodhouse(a)intel.com                              Intel Corporation

[Steve Dickson]

David Woodhouse wrote:
>  On Mon, 2009-01-12 at 16:16 -0500, Josef Bacik wrote:
> > Hello,
> >
> > I've just updated btrfs-progs to 0.17 in rawhide, and will be doing
> > the same for F-10 and F-9 shortly. 
>  
>  Can we have nfs-utils updated too, please? The one we have in F-9
>  predates my patches to use statfs() to obtain a fsid for exporting the
>  file system. Not sure about F-10, but 1.1.4 in rawhide is OK. 
Just to be clear... you are wanting one of or all of the following patch 
back ported to F9?

    Stop exportfs warning about needing fsid, when we actually have one
    Use fsid from statfs for UUID if blkid can't cope (or not used)
    Explicit UUID handling doesn't require blkid; factor out get_uuid_blkdev()
    Fix handling of explicit uuid

steved

[David Woodhouse]

On Fri, 2009-01-16 at 07:52 -0500, Steve Dickson wrote:
>  
>  David Woodhouse wrote:
>  > On Mon, 2009-01-12 at 16:16 -0500, Josef Bacik wrote:
>  >> Hello,
>  >>
>  >> I've just updated btrfs-progs to 0.17 in rawhide, and will be doing
>  >> the same for F-10 and F-9 shortly. 
>  > 
>  > Can we have nfs-utils updated too, please? The one we have in F-9
>  > predates my patches to use statfs() to obtain a fsid for exporting the
>  > file system. Not sure about F-10, but 1.1.4 in rawhide is OK.
>  
>  Just to be clear... you are wanting one of or all of the following patch 
>  back ported to F9?
>  
>      Stop exportfs warning about needing fsid, when we actually have one
>      Use fsid from statfs for UUID if blkid can't cope (or not used)
>      Explicit UUID handling doesn't require blkid; factor out get_uuid_blkdev()
>      Fix handling of explicit uuid 
Those four should be sufficient, I believe. It's dead easy to test --
just export a btrfs file system and try to mount it; see the ESTALE when
the kernel doesn't understand the nfsfh that mountd just gave out.

-- 
David Woodhouse                            Open Source Technology Centre
David.Woodhouse(a)intel.com                              Intel Corporation

[Rakesh Pandit]

2009/1/13 Josef Bacik <josef(a)toxicpanda.com&gt;:
>  Hello,
>  
[..]
>  Btrfs has built in snapshotting, subvolumes and RAID support, so
> lots
>  of fun new things to tinker with.  A few disclaimers 
>  Please feel free to file a bugzilla if something goes horribly
> wrong,
>  stability and finishing out the remaining features is what our main
>  focus is now.  Thanks,
>  
I tried two kernels:

First one:
mkfs.btrfs /dev/sda5

[root@test-rocky /]# uname -a
Linux test-rocky 2.6.29-0.41.rc2.fc11.i686

[root@test-rocky /]# modprobe libcrc32c ; modprobe zlib_inflate
zlib_deflate ; insmod btrfs.ko
FATAL: Module zlib_inflate not found.
insmod: can't read 'btrfs.ko': No such file or directory

btrfsctl -a
Scanning for Btrfs filesystems
cannot register some ioctl error

=========================================

Second one:

mkfs.btrfs /dev/sda5

[root@test-rocky /]# uname -a
Linux test-rocky 2.6.29-0.35.rc1.git4.fc11.i686

btrfsctl -a
Scanning for Btrfs filesystems
failed to open/dev/btrfs-control

I was working on something where I need somewhat stable behaviour, may
you suggest which build I should use ? Or Am I doing something wrong ?

Thanks

-- 
rakesh

[Eric Sandeen]

Rakesh Pandit wrote:
>  2009/1/13 Josef Bacik <josef(a)toxicpanda.com&gt;:
> > Hello,
> >
>  [..]
> > Btrfs has built in snapshotting, subvolumes and RAID support, so lots
> > of fun new things to tinker with.  A few disclaimers
>  
> > Please feel free to file a bugzilla if something goes horribly wrong,
> > stability and finishing out the remaining features is what our main
> > focus is now.  Thanks,
> >
>  
>  I tried two kernels:
>  
>  First one:
>  mkfs.btrfs /dev/sda5
>  
>  [root@test-rocky /]# uname -a
>  Linux test-rocky 2.6.29-0.41.rc2.fc11.i686
>  
>  [root@test-rocky /]# modprobe libcrc32c ; modprobe zlib_inflate
>  zlib_deflate ; insmod btrfs.ko
>  FATAL: Module zlib_inflate not found.
>  insmod: can't read 'btrfs.ko': No such file or directory 
# modprobe btrfs

should suffice

heck just

# mount -t btrfs /dev/sda5 /mnt/blah

should suffice, or on uptodate rawhide, just

# mount /dev/sda5 /mnt/blah

-Eric

[Rakesh Pandit]

2009/1/17 Eric Sandeen <sandeen(a)redhat.com&gt;:
[..]
>  # modprobe btrfs
>
>  should suffice
>
>  heck just
>
>  # mount -t btrfs /dev/sda5 /mnt/blah
>
>  should suffice, or on uptodate rawhide, just
>
>  # mount /dev/sda5 /mnt/blah
>  
Thanks

just small session on mounted partition (/data1 folder):

[root@test-rocky data1]# mkdir dir1
[root@test-rocky data1]# cd dir1/
[root@test-rocky dir1]# touch file1 file2; echo "test" > file3
[root@test-rocky data1]# cd ..

root@test-rocky data1]# btrfsctl -s dir1_snap dir1
ioctl:: Inappropriate ioctl for device

[root@test-rocky data1]# cd ~
[root@test-rocky ~]# btrfsctl -a
Scanning for Btrfs filesystems
failed to read /dev/sr0

I am confused, any suggestions here ?

[root@test-rocky ~]# uname -a
Linux test-rocky 2.6.29-0.35.rc1.git4.fc11.i686

-- 
Regards,
Rakesh Pandit

[Eric Sandeen]

Rakesh Pandit wrote:

>  Thanks
>  
>  just small session on mounted partition (/data1 folder):
>  
>  [root@test-rocky data1]# mkdir dir1
>  [root@test-rocky data1]# cd dir1/
>  [root@test-rocky dir1]# touch file1 file2; echo "test" > file3
>  [root@test-rocky data1]# cd ..
>  
>  root@test-rocky data1]# btrfsctl -s dir1_snap dir1
>  ioctl:: Inappropriate ioctl for device
>  
>  [root@test-rocky data1]# cd ~
>  [root@test-rocky ~]# btrfsctl -a
>  Scanning for Btrfs filesystems
>  failed to read /dev/sr0
>  
>  I am confused, any suggestions here ? 
I'd suggest taking it to the btrfs mailing list :)

-Eric

[Josef Bacik]

On Sat, Jan 17, 2009 at 1:19 PM, Eric Sandeen <sandeen(a)redhat.com&gt; wrote:
>  Rakesh Pandit wrote:
>
> > Thanks
> >
> > just small session on mounted partition (/data1 folder):
> >
> > [root@test-rocky data1]# mkdir dir1
> > [root@test-rocky data1]# cd dir1/
> > [root@test-rocky dir1]# touch file1 file2; echo "test" > file3
> > [root@test-rocky data1]# cd ..
> >
> > root@test-rocky data1]# btrfsctl -s dir1_snap dir1
> > ioctl:: Inappropriate ioctl for device
> >
> > [root@test-rocky data1]# cd ~
> > [root@test-rocky ~]# btrfsctl -a
> > Scanning for Btrfs filesystems
> > failed to read /dev/sr0
> > 
Updated ioctl's for -rc2.  I just pushed the -0.18 package into
rawhide, it should start working tomorrow.  The joys of a devel fs :).

Josef

[Kevin Kofler]

One question I couldn't find a definite answer to: Does btrfs-convert also
work on ext4? In other words: Can we upgrade from ext3 to ext4 and then
from ext4 to btrfs? Or are we painting ourselves in a corner by using ext4?

        Kevin Kofler

[Rahul Sundaram]

Kevin Kofler wrote:
>  One question I couldn't find a definite answer to: Does
> btrfs-convert also
>  work on ext4? In other words: Can we upgrade from ext3 to ext4 and then
>  from ext4 to btrfs? Or are we painting ourselves in a corner by using ext4? 
I asked this question to Eric Sandeen earlier on IRC and Chris Mason has 
confirmed that it should work. Someone probably needs to test that however.

Besides, Btrfs is probably going to take a couple more releases before 
it gets to the stage where we can trust it enough to use it as default 
and Ext4 provides us good advantages meanwhile.

Rahul

[Eric Sandeen]

Rahul Sundaram wrote:
>  Kevin Kofler wrote:
> > One question I couldn't find a definite answer to: Does btrfs-convert also
> > work on ext4? In other words: Can we upgrade from ext3 to ext4 and then
> > from ext4 to btrfs? Or are we painting ourselves in a corner by using ext4?
>  
>  I asked this question to Eric Sandeen earlier on IRC and Chris Mason has 
>  confirmed that it should work. Someone probably needs to test that however.
>  
>  Besides, Btrfs is probably going to take a couple more releases before 
>  it gets to the stage where we can trust it enough to use it as default 
>  and Ext4 provides us good advantages meanwhile. 
It does basically work; the ext4 snapshot is bloated because the tool
doesn't understand uninitialized ext4 block groups, but that's just a
detail.  (But yes, I did test this).

Because the tool was properly written to use libext2fs, ext4 should
"just work" (modulo some little kinks like above).  :)

-Eric