"not enough space" with "data raid0"

"not enough space" with "data raid0"

[Helmut Hullen]

Hallo, linux-btrfs,

I've (once more) created my test system:

  mkfs.btrfs -d raid0 -m raid1 /dev/sdb1 /dev/sdc1

73 GB + 146 GB.

Then I mounted it and copied about 150 GByte onto it. But copying was  
incomplete, the job ended with "no space on ..."


# btrfs fi show

Label: 'Scsi'  uuid: e30586e9-a903-4d17-8ec0-1781457212c6
	Total devices 2 FS bytes used 134.86GB
	devid    1 size 68.37GB used 68.37GB path /dev/sdb1
	devid    2 size 136.73GB used 68.35GB path /dev/sdc1

Btrfs Btrfs v0.19

# btrfs fi df <mountpoint>

Data, RAID0: total=134.68GB, used=134.67GB
Data: total=8.00MB, used=8.00MB
System, RAID1: total=8.00MB, used=16.00KB
System: total=4.00MB, used=0.00
Metadata, RAID1: total=1.00GB, used=185.39MB
Metadata: total=8.00MB, used=0.00

# df <mountpoint>

Filesystem     1K-blocks      Used Available Use% Mounted on
/dev/sdc1      215059324 141597364      8724 100% /mnt/btr

# fdisk -l

Disk /dev/sdb: 73.4 GB, 73407868928 bytes
Disk /dev/sdc: 146.8 GB, 146814976000 bytes

-----------------------------------

Where is the problem, how can I use the full space?

Viele Gruesse!
Helmut

Re: "not enough space" with "data raid0"

[Chris Samuel]

[-- Attachment #1: Type: Text/Plain, Size: 363 bytes --]

On Saturday 17 March 2012 23:01:00 Helmut Hullen wrote:

> Where is the problem, how can I use the full space?

Which kernel was this with Helmut?

cheers,
Chris
-- 
 Chris Samuel  :  http://www.csamuel.org/  :  Melbourne, VIC

This email may come with a PGP signature as a file. Do not panic.
For more info see: http://en.wikipedia.org/wiki/OpenPGP

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 482 bytes --]

Re: "not enough space" with "data raid0"

[Helmut Hullen]

Hallo, Chris,

Du meintest am 17.03.12:

>> Where is the problem, how can I use the full space?

> Which kernel was this with Helmut?

Kernel 3.2.9 (self made)

btrfs-progs-20111030

Viele Gruesse!
Helmut

Re: "not enough space" with "data raid0"

[Hugo Mills]

[-- Attachment #1: Type: text/plain, Size: 1932 bytes --]

On Sat, Mar 17, 2012 at 01:01:00PM +0100, Helmut Hullen wrote:
> Hallo, linux-btrfs,
> 
> I've (once more) created my test system:
> 
>   mkfs.btrfs -d raid0 -m raid1 /dev/sdb1 /dev/sdc1
> 
> 73 GB + 146 GB.
> 
> Then I mounted it and copied about 150 GByte onto it. But copying was  
> incomplete, the job ended with "no space on ..."
> 
> 
> # btrfs fi show
> 
> Label: 'Scsi'  uuid: e30586e9-a903-4d17-8ec0-1781457212c6
> 	Total devices 2 FS bytes used 134.86GB
> 	devid    1 size 68.37GB used 68.37GB path /dev/sdb1
> 	devid    2 size 136.73GB used 68.35GB path /dev/sdc1
> 
> Btrfs Btrfs v0.19
> 
> # btrfs fi df <mountpoint>
> 
> Data, RAID0: total=134.68GB, used=134.67GB
> Data: total=8.00MB, used=8.00MB
> System, RAID1: total=8.00MB, used=16.00KB
> System: total=4.00MB, used=0.00
> Metadata, RAID1: total=1.00GB, used=185.39MB
> Metadata: total=8.00MB, used=0.00
> 
> # df <mountpoint>
> 
> Filesystem     1K-blocks      Used Available Use% Mounted on
> /dev/sdc1      215059324 141597364      8724 100% /mnt/btr
> 
> # fdisk -l
> 
> Disk /dev/sdb: 73.4 GB, 73407868928 bytes
> Disk /dev/sdc: 146.8 GB, 146814976000 bytes
> 
> -----------------------------------
> 
> Where is the problem, how can I use the full space?

   You can't. btrfs requires RAID-0 to be at least two devices wide
(otherwise it's not striped at all, which is the point of RAID-0). If
you want to use the full capacity of both disks and don't care about
the performance gain from striping, use -d single (which is the
default). If you do care about the performance gain from striping,
then you're going to have to lose some usable 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
  --- I am but mad north-north-west:  when the wind is southerly, I ---  
                       know a hawk from a handsaw.                       

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 190 bytes --]

Re: "not enough space" with "data raid0"

[Phillip Susi]

On 3/17/2012 8:19 AM, Hugo Mills wrote:
>> Where is the problem, how can I use the full space?
>
>     You can't. btrfs requires RAID-0 to be at least two devices wide
> (otherwise it's not striped at all, which is the point of RAID-0). If
> you want to use the full capacity of both disks and don't care about
> the performance gain from striping, use -d single (which is the
> default). If you do care about the performance gain from striping,
> then you're going to have to lose some usable space.

So currently btrfs's concept of raid0 is "stripe across as many disks as 
possible, with a minimum of 2 disks".  Is there any reason for that 
minimum?  I don't see why it can't allow only one if that's the best it 
can manage.

Re: "not enough space" with "data raid0"

[Alex]

Phillip Susi <psusi <at> ubuntu.com> writes:

> 
> So currently btrfs's concept of raid0 is "stripe across as many disks as 
> possible, with a minimum of 2 disks".  Is there any reason for that 
> minimum?  I don't see why it can't allow only one if that's the best it 
> can manage.
> 
That's called "Single". http://btrfs.ipv5.de/index.php?title=Mkfs.btrfs

Re: "not enough space" with "data raid0"

[Duncan]

Alex posted on Wed, 28 Mar 2012 22:11:01 +0000 as excerpted:

> Phillip Susi <psusi <at> ubuntu.com> writes:
> 
> 
>> So currently btrfs's concept of raid0 is "stripe across as many disks
>> as possible, with a minimum of 2 disks".  Is there any reason for that
>> minimum?  I don't see why it can't allow only one if that's the best it
>> can manage.
>> 
> That's called "Single". http://btrfs.ipv5.de/index.php?title=Mkfs.btrfs

It's called "single" if you deliberately create it that way, yes.  

However, Phillip's point was, I believe, that if a striped set can start 
with say five devices of varying size and drop one at a time as the 
smallest devices run out of space, there's no reason it should stop at 
two devices.  If the goal is maximum performance, it should insist on 
using only the space available in parallel across all devices and should 
refuse to drop even a single one.  If OTOH, the goal is maximum 
utilization of available space, with a secondary goal of performance if 
at all possible, then it should drop devices all the way down to a single 
device, not stopping at two.

Since btrfs apparently is already content to drop from say five devices 
to two despite the performance drop on the way, there's no fundamental 
reason it should stop there; it should continue all the way down to a 
single device, even tho when it gets to that point it's no longer 
actually striped.

Of course, that's defining the behavior I'd expect of a mature btrfs.  
Right now, it's still experimental, with many missing features.  If at 
this point there's and arbitrary line drawn at two devices minimum, 
that's just the way it is.  But I'd not expect that line to be there when 
the filesystem is declared mature and fully ready for use.

-- 
Duncan - List replies preferred.   No HTML msgs.
"Every nonfree program has a lord, a master --
and if you use the program, he is your master."  Richard Stallman

Re: "not enough space" with "data raid0"

[Alex]

Helmut Hullen <Hullen <at> t-online.de> writes:

> 

> 
> Where is the problem, how can I use the full space?
> 
> Viele Gruesse!
> Helmut

Effectively it's missing the trigger to rebalance when the 'primary' device
starts to get full, or just to randomly spread the data between the devices.

Isn't this going to be a problem for anyone restoring from a backup? Lots of
data heading to btrfs without a pause.

Helmut, can you tell everyone which version of the kernel & patches you're
running pls.

Re: "not enough space" with "data raid0"

[Hugo Mills]

[-- Attachment #1: Type: text/plain, Size: 1031 bytes --]

On Sat, Mar 17, 2012 at 12:19:54PM +0000, Alex wrote:
> Helmut Hullen <Hullen <at> t-online.de> writes:
> 
> > 
> 
> > 
> > Where is the problem, how can I use the full space?
> > 
> > Viele Gruesse!
> > Helmut
> 
> Effectively it's missing the trigger to rebalance when the 'primary' device
> starts to get full, or just to randomly spread the data between the devices.

   No, a balance isn't going to help here. RAID-0 requires a minimum
of 2 chunks in a block group. With two disks, you're only going to be
able to fill the smallest one before you run out of space.

> Isn't this going to be a problem for anyone restoring from a backup? Lots of
> data heading to btrfs without a pause.

   Again, not the issue here.

   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
  --- I am but mad north-north-west:  when the wind is southerly, I ---  
                       know a hawk from a handsaw.                       

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 190 bytes --]

Re: "not enough space" with "data raid0"

[Alex]

> 
>    No, a balance isn't going to help here. RAID-0 requires a minimum
> of 2 chunks in a block group. With two disks, you're only going to be
> able to fill the smallest one before you run out of space.
> 
> > Isn't this going to be a problem for anyone restoring from a backup? Lots of
> > data heading to btrfs without a pause.
> 
>    Again, not the issue here.
> 
>    Hugo.
> 

Thank you HH for the explanation.

What's the 'solution' though to Hugo's situation?
By 'solution' I mean the highest-utility way of dealing with unequal devices
problem in a two or more -up setting.

(I understand that we're violating the expectations of RAID0 and have read the
http://btrfs.ipv5.de/index.php?title=FAQ#How_much_space_do_I_get_with_unequal_devices_in_RAID-1_mode.3F
about RAID1.)

Disks are (were) getting cheaper per MB faster than I could run out of space! So
I have increasing size array of drives - apparently the similar situation as
Hugo (I have three btrfs-ready drives).

I'm struggling how to mkfs.btrfs with this new knowledge whilst resolving it
with the btrfs storage pool concept.

I stress there is a difficulty with my understanding not btrfs.

Kind regards.

Al.

Re: "not enough space" with "data raid0"

[Hugo Mills]

[-- Attachment #1: Type: text/plain, Size: 2135 bytes --]

On Sat, Mar 17, 2012 at 01:36:04PM +0000, Alex wrote:
> >    No, a balance isn't going to help here. RAID-0 requires a minimum
> > of 2 chunks in a block group. With two disks, you're only going to be
> > able to fill the smallest one before you run out of space.
> > 
> > > Isn't this going to be a problem for anyone restoring from a backup? Lots of
> > > data heading to btrfs without a pause.
> > 
> >    Again, not the issue here.
> 
> Thank you HH for the explanation.
> 
> What's the 'solution' though to Hugo's situation?
> By 'solution' I mean the highest-utility way of dealing with unequal devices
> problem in a two or more -up setting.

   Use mkfs.btrfs -d single, as I said in another part of this thread.

   We have a bit of a problem right now with adding a second device to
an existing FS with -d single, where the FS "upgrades" the single to
RAID-0. However, there's patches from Ilya that stop that behaviour,
and the restriper can be used to switch back to single after the
"upgrade".

> (I understand that we're violating the expectations of RAID0 and
> have read the
> http://btrfs.ipv5.de/index.php?title=FAQ#How_much_space_do_I_get_with_unequal_devices_in_RAID-1_mode.3F
> about RAID1.)

   I'm working on a comprehensive explanation of the limits on btrfs's
space usage, and a little JavaScript tool to help plan/explain disk
usage.

> Disks are (were) getting cheaper per MB faster than I could run out
> of space! So I have increasing size array of drives - apparently the
> similar situation as Hugo (I have three btrfs-ready drives).

   I don't have this problem personally. I just know something about
the way btrfs allocates chunks. I think you meant Helmut has the
problem. :)

> I'm struggling how to mkfs.btrfs with this new knowledge whilst resolving it
> with the btrfs storage pool concept.
> 
> I stress there is a difficulty with my understanding not btrfs.

   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
        --- UNIX: Spanish manufacturer of fire extinguishers. ---        

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 190 bytes --]

Re: "not enough space" with "data raid0"

[Helmut Hullen]

Hallo, Hugo,

Du meintest am 17.03.12:

>> What's the 'solution' though to Hugo's situation?
>> By 'solution' I mean the highest-utility way of dealing with unequal
>> devices problem in a two or more -up setting.

>    Use mkfs.btrfs -d single, as I said in another part of this
> thread.

Does "single" allow adding new (bigger) disks and removing old (smaller)  
disks?

Viele Gruesse!
Helmut

Re: "not enough space" with "data raid0"

[Hugo Mills]

[-- Attachment #1: Type: text/plain, Size: 1052 bytes --]

On Sat, Mar 17, 2012 at 03:24:00PM +0100, Helmut Hullen wrote:
> Hallo, Hugo,
> 
> Du meintest am 17.03.12:
> 
> >> What's the 'solution' though to Hugo's situation?
> >> By 'solution' I mean the highest-utility way of dealing with unequal
> >> devices problem in a two or more -up setting.
> 
> >    Use mkfs.btrfs -d single, as I said in another part of this
> > thread.
> 
> Does "single" allow adding new (bigger) disks and removing old (smaller)  
> disks?

   Yes. "single" is effectively "RAID linear" -- just adding the disks
together, effectively linearly(*). It's called "single" because it's
single (i.e. no) redundancy.

   Hugo.

(*) actually the allocator will kind of "stripe" the data in 1GiB-wide
stripes, but that's incidental really.

-- 
=== 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
   --- A diverse working environment:  Di longer you vork here, di ---   
                             verse it gets.                              

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 190 bytes --]

Re: "not enough space" with "data raid0"

[Alex]

Hugo Mills <hugo <at> carfax.org.uk> writes:

> 
> On Sat, Mar 17, 2012 at 01:36:04PM +0000, Alex wrote:

> 
>    I don't have this problem personally. I just know something about
> the way btrfs allocates chunks. I think you meant Helmut has the
> problem. :)
> 
> > I'm struggling how to mkfs.btrfs with this new knowledge whilst resolving it
> > with the btrfs storage pool concept.
> > 
> > I stress there is a difficulty with my understanding not btrfs.
> 
>    Hugo.
> 


Yes, I meant Hugo .. damn .. Helmut, sorry! However deed-poll is a cost
effective measure which would allow you to avoid the coincident first letter
with Helmut. ;-)

Since we're still experimental in the non-SUSE and Oracle world and the space
meta demands sb lower can I squeak by:

mkfs.btrfs -d single -m raid1 ?

Don't laugh.

Re: "not enough space" with "data raid0"

[Alex]

>    I'm working on a comprehensive explanation of the limits on btrfs's
> space usage, and a little JavaScript tool to help plan/explain disk
> usage.

Thank you Hugo.

Given the kernel 3.3 release yesterday [1] we've gained the ability to restripe
to different raid levels after the fact. So I guess I don't have too much to
worry about with decent backups.

The captcha word for this post is "baptizes". Quite apt (no pun intended) really!
 

[1] http://thread.gmane.org/gmane.linux.kernel/1240436

Re: "not enough space" with "data raid0"

[Hugo Mills]

[-- Attachment #1: Type: text/plain, Size: 976 bytes --]

On Mon, Mar 19, 2012 at 03:14:19PM +0000, Alex wrote:
> 
> >    I'm working on a comprehensive explanation of the limits on btrfs's
> > space usage, and a little JavaScript tool to help plan/explain disk
> > usage.
> 
> Thank you Hugo.
> 
> Given the kernel 3.3 release yesterday [1] we've gained the ability to restripe
> to different raid levels after the fact. So I guess I don't have too much to
> worry about with decent backups.

   Yes, you do. RAID is not a backup: What happens if you accidentally
write zeroes over all your superblocks? Or rm -rf /home? RAID will not
help you here.

   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
   --- In the future, terrorists won't be carrying their ID cards. ---   
                   They'll be carrying yours. -- Henry                   
                         Porter,  Suspect Nation                         

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 190 bytes --]

Re: "not enough space" with "data raid0"

[Alex Plumbley-Jones]
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Re: "not enough space" with "data raid0"

[Helmut Hullen]

Hallo, Hugo,

Du meintest am 17.03.12:

[no space left on device ...]

>>> Where is the problem, how can I use the full space?

>> Effectively it's missing the trigger to rebalance when the 'primary'
>> device starts to get full, or just to randomly spread the data
>> between the devices.

>    No, a balance isn't going to help here. RAID-0 requires a minimum
> of 2 chunks in a block group. With two disks, you're only going to be
> able to fill the smallest one before you run out of space.

Ok - it happens only with 2 disks/Partitions?

I've continued playing; added a 3rd partition/device and then balanced:


# btrfs device add /dev/sdd1 /mnt/btr
## 73 + 146 + 146 GByte

# df

Filesystem             1K-blocks       Used  Available Use% Mounted on
/dev/sdc1              358432300  225400136   59842800  80% /mnt/btr

# added about 70 GByte (all together more than 3 times the smallest partition)

# btrfs fi show

Label: 'Scsi'  uuid: e30586e9-a903-4d17-8ec0-1781457212c6
	Total devices 3 FS bytes used 214.67GB
	devid    1 size 68.37GB used 68.37GB path /dev/sdb1
	devid    3 size 136.73GB used 41.01GB path /dev/sdd1
	devid    2 size 136.73GB used 109.36GB path /dev/sdc1

Btrfs Btrfs v0.19

# btrfs fi df /mnt/btr

Data, RAID0: total=216.71GB, used=214.38GB
System, RAID1: total=8.00MB, used=24.00KB
System: total=4.00MB, used=0.00
Metadata, RAID1: total=1.00GB, used=295.13MB

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

# btrfs fi balance /mnt/btr

# btrfs fi show

Label: 'Scsi'  uuid: e30586e9-a903-4d17-8ec0-1781457212c6
	Total devices 3 FS bytes used 214.67GB
	devid    1 size 68.37GB used 67.34GB path /dev/sdb1
	devid    3 size 136.73GB used 40.85GB path /dev/sdd1
	devid    2 size 136.73GB used 107.85GB path /dev/sdc1

Btrfs Btrfs v0.19

# btrfs fi df /mnt/btr

Data, RAID0: total=215.01GB, used=214.38GB
System, RAID1: total=8.00MB, used=24.00KB
System: total=4.00MB, used=0.00
Metadata, RAID1: total=512.00MB, used=294.88MB

----------------------------------------------------------------------

Looks as desired, the 3-disks-system contains more than 3 times the  
smallest disk.

Balancing hasn't redistributed the contents - no problem.

By the way: you should name the prefixes in the NIST way for powers of  
2: KiB, MiB, GiB. Or change to decimal prefixes.

Viele Gruesse!
Helmut

Re: "not enough space" with "data raid0"

[Hugo Mills]

[-- Attachment #1: Type: text/plain, Size: 1816 bytes --]

On Sat, Mar 17, 2012 at 02:46:00PM +0100, Helmut Hullen wrote:
> Hallo, Hugo,
> 
> Du meintest am 17.03.12:
> 
> [no space left on device ...]
> 
> >>> Where is the problem, how can I use the full space?
> 
> >> Effectively it's missing the trigger to rebalance when the 'primary'
> >> device starts to get full, or just to randomly spread the data
> >> between the devices.
> 
> >    No, a balance isn't going to help here. RAID-0 requires a minimum
> > of 2 chunks in a block group. With two disks, you're only going to be
> > able to fill the smallest one before you run out of space.
> 
> Ok - it happens only with 2 disks/Partitions?

   Not quite. With the current RAID-0 implementation, you will lose
the difference between the largest disk and the second-largest.
There's a proposal from November that may help with that, but I
haven't had a chance to look at it yet to see what the implications
are.

> I've continued playing; added a 3rd partition/device and then balanced:
> 
> 
> # btrfs device add /dev/sdd1 /mnt/btr
> ## 73 + 146 + 146 GByte

   OK, so in this case the largest and second-largest devices are the
same, so you will lose no space.

[snip]

> Looks as desired, the 3-disks-system contains more than 3 times the  
> smallest disk.
> 
> Balancing hasn't redistributed the contents - no problem.
> 
> By the way: you should name the prefixes in the NIST way for powers of  
> 2: KiB, MiB, GiB. Or change to decimal prefixes.

   I know. It's something I submitted patches for a long time ago, and
they never got taken up. It's on my list of things to fix.

   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
        --- UNIX: Spanish manufacturer of fire extinguishers. ---        

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 190 bytes --]