unexpected raid1 behavior?

unexpected raid1 behavior?

[Jean-Philippe Robichaud]

Hi, I'm starting to play with btrfs on my new computer. I'm running Gentoo and 
have compiled the 2.6.31 kernel, enabling btrfs.

Now I have 2 partitions (on 2 different sata disks) that are free for me to 
play with, each about 375 gb in size.  I wanted to create a "raid1" volume 
using these two partitions, so I did:
 
# mkfs.btrfs -d raid1 /dev/sda5 /dev/sdb5
# mount /dev/sda5 /btrfs

and everything seems fine.

Now what I find strange is that everything looks like a raid0 was created, not 
a raid1:

$ df -h /btrfs/
 Filesystem            Size  Used Avail Use% Mounted on
/dev/sda5             684G   72G  612G  11% /btrfs

What am I doing (or understanding) wrong?

Thanks!

Jp

Re: unexpected raid1 behavior?

[Hugo Mills]

[-- Attachment #1: Type: text/plain, Size: 1484 bytes --]

yOn Fri, Nov 27, 2009 at 03:41:17PM -0500, Jean-Philippe Robichaud wrote:
> Now I have 2 partitions (on 2 different sata disks) that are free for me to 
> play with, each about 375 gb in size.  I wanted to create a "raid1" volume 
> using these two partitions, so I did:
>  
> # mkfs.btrfs -d raid1 /dev/sda5 /dev/sdb5
> # mount /dev/sda5 /btrfs
> 
> and everything seems fine.
> 
> Now what I find strange is that everything looks like a raid0 was created, not 
> a raid1:
> 
> $ df -h /btrfs/
>  Filesystem            Size  Used Avail Use% Mounted on
> /dev/sda5             684G   72G  612G  11% /btrfs
> 
> What am I doing (or understanding) wrong?

   It's effectively showing you the number of unallocated blocks, so
(with a RAID-1, single-redundancy filesystem), files will appear to
take twice as much of your free space as you think they should: write
a 2GiB file to that filesystem, and free space will drop by 4GiB.

   I think that the reasoning behind this is that if you're using
per-object mirroring/striping, it's impossible to give a precise count
of the free space remaining on the volume in any meaningful way: write
a 1GiB striped file, and you'll take 1GiB of space; write the same
file mirrored, and you'll take 2GiB of space.

   Hugo.

-- 
=== Hugo Mills: hugo@... carfax.org.uk | darksatanic.net | lug.org.uk ===
  PGP key: 515C238D from wwwkeys.eu.pgp.net or http://www.carfax.org.uk
           --- There are three mistaiks in this sentance. ---            

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: unexpected raid1 behavior?

[Jean-Philippe Robichaud]

On Friday 27 November 2009 16:04:42 Hugo Mills wrote:
> yOn Fri, Nov 27, 2009 at 03:41:17PM -0500, Jean-Philippe Robichaud wrote:
> > Now I have 2 partitions (on 2 different sata disks) that are free for me
> > to play with, each about 375 gb in size.  I wanted to create a "raid1"
> > volume using these two partitions, so I did:
> >
> > # mkfs.btrfs -d raid1 /dev/sda5 /dev/sdb5
> > # mount /dev/sda5 /btrfs
> >
> > and everything seems fine.
> >
> > Now what I find strange is that everything looks like a raid0 was
> > created, not a raid1:
> >
> > $ df -h /btrfs/
> >  Filesystem            Size  Used Avail Use% Mounted on
> > /dev/sda5             684G   72G  612G  11% /btrfs
> >
> > What am I doing (or understanding) wrong?
> 
>    It's effectively showing you the number of unallocated blocks, so
> (with a RAID-1, single-redundancy filesystem), files will appear to
> take twice as much of your free space as you think they should: write
> a 2GiB file to that filesystem, and free space will drop by 4GiB.
> 
>    I think that the reasoning behind this is that if you're using
> per-object mirroring/striping, it's impossible to give a precise count
> of the free space remaining on the volume in any meaningful way: write
> a 1GiB striped file, and you'll take 1GiB of space; write the same
> file mirrored, and you'll take 2GiB of space.
> 
>    Hugo.
> 
Thanks for your input Hugo, but somehow it looks like something different is 
happening:

du toto
1088 toto
$ df .
/dev/sda5            716579256  75455900 641123356  11% /btrfs
$ cp toto toto2 
$ btrfsctl -c .
$ df .
/dev/sda5            716579256  75456992 641122264  11% /btrfs

So 1092 block were 'consumed', we're far from the +2000 I would have 
expected...

Is there a place in /sys or /proc where I could perhaps get 'stats' about the 
btrfs volume?

Re: unexpected raid1 behavior?

[Sander]

Hello Jean-Philippe,

Jean-Philippe Robichaud wrote (ao):
> Is there a place in /sys or /proc where I could perhaps get 'stats'
> about the btrfs volume?

I think btrfs-show should help you there.

	With kind regards, Sander

-- 
Humilis IT Services and Solutions
http://www.humilis.net

Re: unexpected raid1 behavior?

[Jean-Philippe Robichaud]

On Sunday 29 November 2009 07:42:56 Sander wrote:
> Hello Jean-Philippe,
> 
> Jean-Philippe Robichaud wrote (ao):
> > Is there a place in /sys or /proc where I could perhaps get 'stats'
> > about the btrfs volume?
> 
> I think btrfs-show should help you there.
> 
> 	With kind regards, Sander
> 
Sure: here it goes:

# sudo btrfs-show
failed to read /dev/sr0

Label: none  uuid: fb9270b4-8779-489f-a599-eaab299d4633
        Total devices 2 FS bytes used 78.98GB
        devid    1 size 341.69GB used 146.03GB path /dev/sda5
        devid    2 size 341.69GB used 146.01GB path /dev/sdb5

Re: unexpected raid1 behavior?

[jean-philippe robichaud]

On Sun, Nov 29, 2009 at 4:28 PM, Jean-Philippe Robichaud
<jean.philippe.robichaud@gmail.com> wrote:
>
> On Sunday 29 November 2009 07:42:56 Sander wrote:
> > Hello Jean-Philippe,
> >
> > Jean-Philippe Robichaud wrote (ao):
> > > Is there a place in /sys or /proc where I could perhaps get 'stat=
s'
> > > about the btrfs volume?
> >
> > I think btrfs-show should help you there.
> >
> > =A0 =A0 =A0 With kind regards, Sander
> >
> Sure: here it goes:
>
> # sudo btrfs-show
> failed to read /dev/sr0
>
> Label: none =A0uuid: fb9270b4-8779-489f-a599-eaab299d4633
> =A0 =A0 =A0 =A0Total devices 2 FS bytes used 78.98GB
> =A0 =A0 =A0 =A0devid =A0 =A01 size 341.69GB used 146.03GB path /dev/s=
da5
> =A0 =A0 =A0 =A0devid =A0 =A02 size 341.69GB used 146.01GB path /dev/s=
db5
>

I'm not sure I actually understand how to read this.The size of each
device make sense (341GB) as it is the size of both partitions.  Now
the FS-bytes-used also make sense as it is about the size of data that
I've copied, but the per-device used space seems strange to me and the
output of "df -h" is also hard to interpret:

=46ilesystem            Size  Used Avail Use% Mounted on
/dev/sda5             684G   79G  605G  12% /btrfs

I would have understood if the "Used" field would have been 160GB or
if "Avail" would have dropped to 520GB.

I guess that I'll come to a point where Used will be 340GB and Avail
will also be 340GB but I'll be getting No Space Left On Device errors?
--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs" =
in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html