Same size drive has less usable space

Same size drive has less usable space

[Luciano ES]

I have a new drive for backups. 
I copied everything over and now I have this problem:

Filesystem      Size  Used 	 Avail 	Use% 	Mounted on
/dev/sda1   	931G  914G   18G  	99% 	/xx
/dev/sdb1   	931G  920G   11G  	99% 	/xxbkp

So 914GB from the old drive expand and become 920GB. The new drive 
is supposed to be the same size, but for some reason it can't 
really hold it all. I will be forced to waste precious gigabytes.

I tried to format the new one exactly like the old one, but 
it was not possible:

$ xfs_info /xx
meta-data=/dev/sda1     isize=256    agcount=8, agsize=30506944 blks
         =              sectsz=4096  attr=2, projid32bit=1
         =              crc=0        finobt=0 spinodes=0 rmapbt=0
         =              reflink=0
data     =              bsize=4096   blocks=244055552, imaxpct=25
         =              sunit=0      swidth=0 blks
naming   =version 2     bsize=4096   ascii-ci=0 ftype=0
log      =internal      bsize=4096   blocks=119167, version=2
         =              sectsz=4096  sunit=1 blks, lazy-count=1
realtime =none          extsz=4096   blocks=0, rtextents=0


$ xfs_info /xxbkp
meta-data=/dev/sdb1     isize=512    agcount=8, agsize=61013888 blks
         =              sectsz=512   attr=2, projid32bit=1
         =              crc=1        finobt=1 spinodes=1 rmapbt=0
         =              reflink=0
data     =              bsize=2048   blocks=488111104, imaxpct=25
         =              sunit=0      swidth=0 blks
naming   =version 2     bsize=4096   ascii-ci=0 ftype=1
log      =internal      bsize=2048   blocks=238335, version=2
         =              sectsz=512   sunit=0 blks, lazy-count=1
realtime =none          extsz=4096   blocks=0, rtextents=0


I guess at least part of the problem is CRC enabled in the second 
one. So, is there anything I can do to make all the data fit in the 
new drive?

TIA

-- 
Luciano ES
>>

Re: Same size drive has less usable space

[Carlos E. R.]

[-- Attachment #1.1: Type: text/plain, Size: 477 bytes --]

On 20/12/2018 16.13, Luciano ES wrote:
> I have a new drive for backups. 
> I copied everything over and now I have this problem:

...

> I guess at least part of the problem is CRC enabled in the second 
> one. So, is there anything I can do to make all the data fit in the 
> new drive?

Clone it... dd, for instance, but then you have the problem of changing
the uuid.

-- 
Cheers / Saludos,

		Carlos E. R.
		(from 42.3 x86_64 "Malachite" at Telcontar)


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 181 bytes --]

Re: Same size drive has less usable space

[Eric Sandeen]

On 12/20/18 9:13 AM, Luciano ES wrote:
> I have a new drive for backups. 
> I copied everything over and now I have this problem:
> 
> Filesystem      Size  Used 	 Avail 	Use% 	Mounted on
> /dev/sda1   	931G  914G   18G  	99% 	/xx
> /dev/sdb1   	931G  920G   11G  	99% 	/xxbkp
> 
> So 914GB from the old drive expand and become 920GB. The new drive 
> is supposed to be the same size, but for some reason it can't 
> really hold it all. I will be forced to waste precious gigabytes.
> 
> I tried to format the new one exactly like the old one, but 
> it was not possible:
> 
> $ xfs_info /xx
> meta-data=/dev/sda1     isize=256    agcount=8, agsize=30506944 blks
>          =              sectsz=4096  attr=2, projid32bit=1
>          =              crc=0        finobt=0 spinodes=0 rmapbt=0
>          =              reflink=0
> data     =              bsize=4096   blocks=244055552, imaxpct=25
>          =              sunit=0      swidth=0 blks
> naming   =version 2     bsize=4096   ascii-ci=0 ftype=0
> log      =internal      bsize=4096   blocks=119167, version=2
>          =              sectsz=4096  sunit=1 blks, lazy-count=1
> realtime =none          extsz=4096   blocks=0, rtextents=0
> 
> 
> $ xfs_info /xxbkp
> meta-data=/dev/sdb1     isize=512    agcount=8, agsize=61013888 blks
>          =              sectsz=512   attr=2, projid32bit=1
>          =              crc=1        finobt=1 spinodes=1 rmapbt=0
>          =              reflink=0
> data     =              bsize=2048   blocks=488111104, imaxpct=25
>          =              sunit=0      swidth=0 blks
> naming   =version 2     bsize=4096   ascii-ci=0 ftype=1
> log      =internal      bsize=2048   blocks=238335, version=2
>          =              sectsz=512   sunit=0 blks, lazy-count=1
> realtime =none          extsz=4096   blocks=0, rtextents=0
> 
> 
> I guess at least part of the problem is CRC enabled in the second 
> one. So, is there anything I can do to make all the data fit in the 
> new drive?

Well, you have larger inodes on xxbkp for starters.  If you want,

# mkfs.xfs -m crc=0,finobt=1 -i sparse=0 -n ftype=0

should get you the same geometry.  This is all documented in the mfks.xfs
manpage, btw.

But beware that copying sparse files w/o maintaining sparseness (for example)
will also consume more space.  If you want a nothing less than a bit-for-bit
copy, use dd.  ;)

-Eric

Re: Same size drive has less usable space

[Darrick J. Wong]

On Thu, Dec 20, 2018 at 10:10:42AM -0600, Eric Sandeen wrote:
> On 12/20/18 9:13 AM, Luciano ES wrote:
> > I have a new drive for backups. 
> > I copied everything over and now I have this problem:
> > 
> > Filesystem      Size  Used 	 Avail 	Use% 	Mounted on
> > /dev/sda1   	931G  914G   18G  	99% 	/xx
> > /dev/sdb1   	931G  920G   11G  	99% 	/xxbkp
> > 
> > So 914GB from the old drive expand and become 920GB. The new drive 
> > is supposed to be the same size, but for some reason it can't 
> > really hold it all. I will be forced to waste precious gigabytes.
> > 
> > I tried to format the new one exactly like the old one, but 
> > it was not possible:
> > 
> > $ xfs_info /xx
> > meta-data=/dev/sda1     isize=256    agcount=8, agsize=30506944 blks
> >          =              sectsz=4096  attr=2, projid32bit=1
> >          =              crc=0        finobt=0 spinodes=0 rmapbt=0
> >          =              reflink=0
> > data     =              bsize=4096   blocks=244055552, imaxpct=25
> >          =              sunit=0      swidth=0 blks
> > naming   =version 2     bsize=4096   ascii-ci=0 ftype=0
> > log      =internal      bsize=4096   blocks=119167, version=2
> >          =              sectsz=4096  sunit=1 blks, lazy-count=1
> > realtime =none          extsz=4096   blocks=0, rtextents=0
> > 
> > 
> > $ xfs_info /xxbkp
> > meta-data=/dev/sdb1     isize=512    agcount=8, agsize=61013888 blks
> >          =              sectsz=512   attr=2, projid32bit=1
> >          =              crc=1        finobt=1 spinodes=1 rmapbt=0
> >          =              reflink=0
> > data     =              bsize=2048   blocks=488111104, imaxpct=25
> >          =              sunit=0      swidth=0 blks
> > naming   =version 2     bsize=4096   ascii-ci=0 ftype=1
> > log      =internal      bsize=2048   blocks=238335, version=2
> >          =              sectsz=512   sunit=0 blks, lazy-count=1
> > realtime =none          extsz=4096   blocks=0, rtextents=0
> > 
> > 
> > I guess at least part of the problem is CRC enabled in the second 
> > one. So, is there anything I can do to make all the data fit in the 
> > new drive?
> 
> Well, you have larger inodes on xxbkp for starters.  If you want,
> 
> # mkfs.xfs -m crc=0,finobt=1 -i sparse=0 -n ftype=0
> 
> should get you the same geometry.  This is all documented in the mfks.xfs
> manpage, btw.
> 
> But beware that copying sparse files w/o maintaining sparseness (for example)
> will also consume more space.  If you want a nothing less than a bit-for-bit
> copy, use dd.  ;)

Why not xfs_copy?

--D

> -Eric

Re: Same size drive has less usable space

[Luciano ES]

On Thu, 20 Dec 2018 10:10:42 -0600, Eric Sandeen wrote:

> Well, you have larger inodes on xxbkp for starters.

I can't make smaller inodes on this disk.

"specified blocksize 2048 is less than device physical sector size 4096
switching to logical sector size 512"

another attempt:

"specified blocksize 2048 is less than device physical sector size 4096
switching to logical sector size 512
Minimum inode size for CRCs is 512 bytes"

So yes, it's the CRC feature.


On Thu, 20 Dec 2018 10:10:42 -0600, Eric Sandeen wrote:

> # mkfs.xfs -m crc=0,finobt=1 -i sparse=0 -n ftype=0

Your command line doesn't work. It says:

finobt not supported without CRC support

I have confirmed that disabling CRC allows me to have smaller inodes, 
but I wish I could keep that feature. It sounds like something I want 
to have, to improve reliability.

Is there another way?


On Thu, 20 Dec 2018 08:26:41 -0800, Darrick J. Wong wrote:

> Why not xfs_copy?

Good question. But then the copy would not have CRC enabled either 
because the original doesn't. 

Looks like I an going to have to commit to having CRC always on 
from now on. And I am going to lose about 7GB of space for that.

Is it worth it?


-- 
Luciano ES
>>

Re: Same size drive has less usable space

[Eric Sandeen]

On 12/20/18 10:26 AM, Darrick J. Wong wrote:
> On Thu, Dec 20, 2018 at 10:10:42AM -0600, Eric Sandeen wrote:
>> On 12/20/18 9:13 AM, Luciano ES wrote:
>>> I have a new drive for backups. 
>>> I copied everything over and now I have this problem:
>>>
>>> Filesystem      Size  Used 	 Avail 	Use% 	Mounted on
>>> /dev/sda1   	931G  914G   18G  	99% 	/xx
>>> /dev/sdb1   	931G  920G   11G  	99% 	/xxbkp
>>>
>>> So 914GB from the old drive expand and become 920GB. The new drive 
>>> is supposed to be the same size, but for some reason it can't 
>>> really hold it all. I will be forced to waste precious gigabytes.
>>>
>>> I tried to format the new one exactly like the old one, but 
>>> it was not possible:
>>>
>>> $ xfs_info /xx
>>> meta-data=/dev/sda1     isize=256    agcount=8, agsize=30506944 blks
>>>          =              sectsz=4096  attr=2, projid32bit=1
>>>          =              crc=0        finobt=0 spinodes=0 rmapbt=0
>>>          =              reflink=0
>>> data     =              bsize=4096   blocks=244055552, imaxpct=25
>>>          =              sunit=0      swidth=0 blks
>>> naming   =version 2     bsize=4096   ascii-ci=0 ftype=0
>>> log      =internal      bsize=4096   blocks=119167, version=2
>>>          =              sectsz=4096  sunit=1 blks, lazy-count=1
>>> realtime =none          extsz=4096   blocks=0, rtextents=0
>>>
>>>
>>> $ xfs_info /xxbkp
>>> meta-data=/dev/sdb1     isize=512    agcount=8, agsize=61013888 blks
>>>          =              sectsz=512   attr=2, projid32bit=1
>>>          =              crc=1        finobt=1 spinodes=1 rmapbt=0
>>>          =              reflink=0
>>> data     =              bsize=2048   blocks=488111104, imaxpct=25
>>>          =              sunit=0      swidth=0 blks
>>> naming   =version 2     bsize=4096   ascii-ci=0 ftype=1
>>> log      =internal      bsize=2048   blocks=238335, version=2
>>>          =              sectsz=512   sunit=0 blks, lazy-count=1
>>> realtime =none          extsz=4096   blocks=0, rtextents=0
>>>
>>>
>>> I guess at least part of the problem is CRC enabled in the second 
>>> one. So, is there anything I can do to make all the data fit in the 
>>> new drive?
>>
>> Well, you have larger inodes on xxbkp for starters.  If you want,
>>
>> # mkfs.xfs -m crc=0,finobt=1 -i sparse=0 -n ftype=0
>>
>> should get you the same geometry.  This is all documented in the mfks.xfs
>> manpage, btw.
>>
>> But beware that copying sparse files w/o maintaining sparseness (for example)
>> will also consume more space.  If you want a nothing less than a bit-for-bit
>> copy, use dd.  ;)
> 
> Why not xfs_copy?

Or that :)

-Eric

Re: Same size drive has less usable space

[Eric Sandeen]

On 12/20/18 10:55 AM, Luciano ES wrote:
> On Thu, 20 Dec 2018 10:10:42 -0600, Eric Sandeen wrote:
> 
>> Well, you have larger inodes on xxbkp for starters.
> 
> I can't make smaller inodes on this disk.
> 
> "specified blocksize 2048 is less than device physical sector size 4096
> switching to logical sector size 512"

1) that's a warning not an error
2) that's blocksize not inode size

but ok, I missed that you had 2k blocks on your backup disk.... I don't
really know what you're trying to do here, TBH.

> another attempt:
> 
> "specified blocksize 2048 is less than device physical sector size 4096
> switching to logical sector size 512
> Minimum inode size for CRCs is 512 bytes"

So, I don't think you turned off crcs here.

> So yes, it's the CRC feature.

So turn it off.

> 
> On Thu, 20 Dec 2018 10:10:42 -0600, Eric Sandeen wrote:
> 
>> # mkfs.xfs -m crc=0,finobt=1 -i sparse=0 -n ftype=0
> 
> Your command line doesn't work. It says:
> 
> finobt not supported without CRC support

yeah I fat-fingered that, make it finobt=0 of course.

> I have confirmed that disabling CRC allows me to have smaller inodes, 
> but I wish I could keep that feature. It sounds like something I want 
> to have, to improve reliability.
> 
> Is there another way?

You're over-constraining yourself here; crcs require larger inodes which
requires more space, by definition.  Sorry, but you can't get something
for nothing.

If you are trying to consume identical space you're going to need to use
an identical format.

-Eric

Re: Same size drive has less usable space

[Darrick J. Wong]

On Thu, Dec 20, 2018 at 01:13:05PM -0200, Luciano ES wrote:
> I have a new drive for backups. 
> I copied everything over and now I have this problem:
> 
> Filesystem      Size  Used 	 Avail 	Use% 	Mounted on
> /dev/sda1   	931G  914G   18G  	99% 	/xx
> /dev/sdb1   	931G  920G   11G  	99% 	/xxbkp
> 
> So 914GB from the old drive expand and become 920GB. The new drive 
> is supposed to be the same size, but for some reason it can't 
> really hold it all. I will be forced to waste precious gigabytes.
> 
> I tried to format the new one exactly like the old one, but 
> it was not possible:
> 
> $ xfs_info /xx
> meta-data=/dev/sda1     isize=256    agcount=8, agsize=30506944 blks
>          =              sectsz=4096  attr=2, projid32bit=1
>          =              crc=0        finobt=0 spinodes=0 rmapbt=0
>          =              reflink=0
> data     =              bsize=4096   blocks=244055552, imaxpct=25
>          =              sunit=0      swidth=0 blks
> naming   =version 2     bsize=4096   ascii-ci=0 ftype=0
> log      =internal      bsize=4096   blocks=119167, version=2
>          =              sectsz=4096  sunit=1 blks, lazy-count=1
> realtime =none          extsz=4096   blocks=0, rtextents=0
> 
> 
> $ xfs_info /xxbkp
> meta-data=/dev/sdb1     isize=512    agcount=8, agsize=61013888 blks
>          =              sectsz=512   attr=2, projid32bit=1
>          =              crc=1        finobt=1 spinodes=1 rmapbt=0
>          =              reflink=0
> data     =              bsize=2048   blocks=488111104, imaxpct=25

Just FYI you'll get a lot better performance from the filesystem if the
blocksize is the same as cpu page size (4k I think) and if the
filesystem isn't totally full.

But yeah, either unmount /xx and use xfs_copy one disk to the other, or
format the backup drive with the same parameters (I'll leave that to the
other parts of this thread).

--D

>          =              sunit=0      swidth=0 blks
> naming   =version 2     bsize=4096   ascii-ci=0 ftype=1
> log      =internal      bsize=2048   blocks=238335, version=2
>          =              sectsz=512   sunit=0 blks, lazy-count=1
> realtime =none          extsz=4096   blocks=0, rtextents=0
> 
> 
> I guess at least part of the problem is CRC enabled in the second 
> one. So, is there anything I can do to make all the data fit in the 
> new drive?
> 
> TIA
> 
> -- 
> Luciano ES
> >>

Re: Same size drive has less usable space

[Luciano ES]

On Thu, 20 Dec 2018 11:32:33 -0600, Eric Sandeen wrote:

> but ok, I missed that you had 2k blocks on your backup disk.... I
> don't really know what you're trying to do here, TBH.

Smaller blocks waste less space. In ext3/4, I've always used 2k or 
even 1k blocks. It does save space. Doesn't it in XFS too?

-- 
Luciano ES
>>

Re: Same size drive has less usable space

[Carlos E. R.]

[-- Attachment #1.1: Type: text/plain, Size: 795 bytes --]

On 20/12/2018 19.39, Luciano ES wrote:
> On Thu, 20 Dec 2018 11:32:33 -0600, Eric Sandeen wrote:
> 
>> but ok, I missed that you had 2k blocks on your backup disk.... I
>> don't really know what you're trying to do here, TBH.
> 
> Smaller blocks waste less space. In ext3/4, I've always used 2k or 
> even 1k blocks. It does save space. Doesn't it in XFS too?

Depends. Not always "smaller blocks waste less space". It depends on the
mix of filesizes. If most files are big, then no.

Consider, for instance, that in a filesystem with a fixed number of
inodes (like ext2/3/4), when using small blocks the structures that list
the blocks are larger, simply because there are more blocks.

-- 
Cheers / Saludos,

		Carlos E. R.
		(from 42.3 x86_64 "Malachite" at Telcontar)


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 181 bytes --]

Re: Same size drive has less usable space

[Emmanuel Florac]

[-- Attachment #1: Type: text/plain, Size: 668 bytes --]

Le Thu, 20 Dec 2018 13:13:05 -0200
Luciano ES <lucmove@gmail.com> écrivait:

> I guess at least part of the problem is CRC enabled in the second 
> one. So, is there anything I can do to make all the data fit in the 
> new drive?

Your new drive is using 4K blocks, while the old one uses 512b blocks.
That's why they can't match.

-- 
------------------------------------------------------------------------
Emmanuel Florac     |   Direction technique
                    |   Intellique
                    |	<eflorac@intellique.com>
                    |   +33 1 78 94 84 02
------------------------------------------------------------------------

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 163 bytes --]

Re: Same size drive has less usable space

[Luciano ES]

On Fri, 21 Dec 2018 17:57:37 +0100, Emmanuel Florac wrote:

> Le Thu, 20 Dec 2018 13:13:05 -0200
> Luciano ES écrivait:
> 
> > I guess at least part of the problem is CRC enabled in the second 
> > one. So, is there anything I can do to make all the data fit in the 
> > new drive?  
> 
> Your new drive is using 4K blocks, while the old one uses 512b blocks.
> That's why they can't match.
**************************

What 512b? There is no 512 in the old drive.

$ xfs_info /xx
meta-data=/dev/sda1     isize=256    agcount=8, agsize=30506944 blks
         =              sectsz=4096  attr=2, projid32bit=1
         =              crc=0        finobt=0 spinodes=0 rmapbt=0
         =              reflink=0
data     =              bsize=4096   blocks=244055552, imaxpct=25
         =              sunit=0      swidth=0 blks
naming   =version 2     bsize=4096   ascii-ci=0 ftype=0
log      =internal      bsize=4096   blocks=119167, version=2
         =              sectsz=4096  sunit=1 blks, lazy-count=1
realtime =none          extsz=4096   blocks=0, rtextents=0


$ xfs_info /xxbkp
meta-data=/dev/sdb1     isize=512    agcount=8, agsize=61013888 blks
         =              sectsz=512   attr=2, projid32bit=1
         =              crc=1        finobt=1 spinodes=1 rmapbt=0
         =              reflink=0
data     =              bsize=2048   blocks=488111104, imaxpct=25
         =              sunit=0      swidth=0 blks
naming   =version 2     bsize=4096   ascii-ci=0 ftype=1
log      =internal      bsize=2048   blocks=238335, version=2
         =              sectsz=512   sunit=0 blks, lazy-count=1
realtime =none          extsz=4096   blocks=0, rtextents=0

-- 
Luciano ES
>>