mkfs.xfs: cannot (re)set log sunit=0

mkfs.xfs: cannot (re)set log sunit=0

[Mario 'BitKoenig' Holbe]

[-- Attachment #1.1: Type: text/plain, Size: 2108 bytes --]

Hello,

I intend to set up a XFS filesystem on a RAID0 (Linux md) with 512k
Chunk Size. Since I cannot align the log section sunit to 512k, and
since the filesystem will have not much write activity anyways (in fact,
it'll be mounted read-only most of the time), I tried not to align the
log section sunit at all via -l sunit=0, but this does not appear to
work:

# truncate -s 1G /tmp/foo{0,1}
# losetup /dev/loop0 /tmp/foo0; losetup /dev/loop1 /tmp/foo1
# mdadm -C -l raid0 -n 2 /dev/md9 /dev/loop[01]
# mkfs.xfs -l sunit=0 /dev/md9
log stripe unit (524288 bytes) is too large (maximum is 256KiB)
log stripe unit adjusted to 32KiB
meta-data=/dev/md9               isize=256    agcount=8, agsize=65408 blks
         =                       sectsz=512   attr=2
data     =                       bsize=4096   blocks=523264, imaxpct=25
         =                       sunit=128    swidth=256 blks
naming   =version 2              bsize=4096   ascii-ci=0
log      =internal log           bsize=4096   blocks=2560, version=2
         =                       sectsz=512   sunit=8 blks, lazy-count=1
realtime =none                   extsz=4096   blocks=0, rtextents=0

Is this a bug or a feature?

The minimum I was able to limit the log sunit to was -l sunit=8, i.e. 1
block. Is there a difference between sunit=0 and sunit=1 in practice or
are single log entries aligned to blocks anyways?


Btw...
Since the filesystem is mostly read-only, I was also thinking about
reducing the size of the log section. The planned filesystem will be
5.5T or bigger, which results in 2G log section per default. I thought
about limiting it to 128M or 64M. There will be no concurrent write
activity on the fs. Together with the zero-alignment, what do you think
about that?


PS: I'm not on this list, please CC: me in replies. My Mail-Followup-To:
header should take care of that.


Thanks for your help & regards
   Mario
-- 
File names are infinite in length where infinity is set to 255 characters.
                                -- Peter Collinson, "The Unix File System"

[-- Attachment #1.2: Digital signature --]
[-- Type: application/pgp-signature, Size: 482 bytes --]

[-- Attachment #2: Type: text/plain, Size: 121 bytes --]

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

Re: mkfs.xfs: cannot (re)set log sunit=0

[Dave Chinner]

On Mon, Jul 26, 2010 at 12:12:39PM +0200, Mario 'BitKoenig' Holbe wrote:
> Hello,
> 
> I intend to set up a XFS filesystem on a RAID0 (Linux md) with 512k
> Chunk Size. Since I cannot align the log section sunit to 512k, and
> since the filesystem will have not much write activity anyways (in fact,
> it'll be mounted read-only most of the time), I tried not to align the
> log section sunit at all via -l sunit=0, but this does not appear to
> work:

Of course - when you specific a value yourself, it has to be a valid
value and for version 2 logs that minimum value for th lsunit is one
log block.  Given that:

....
> log      =internal log           bsize=4096   blocks=2560, version=2
....
                                   ^^^^^^^^^^
Your log block size is 4k, then:

> Is this a bug or a feature?
> 
> The minimum I was able to limit the log sunit to was -l sunit=8, i.e. 1
> block.

This is the correct minimum you are allowed to specify.

> Is there a difference between sunit=0 and sunit=1 in practice or
> are single log entries aligned to blocks anyways?

sunit=0 is an invalid configuration for a version 2 log. Filesystems
without a data sunit default to a lsunit = 1 log block, so in
practice version 2 logs are always aligned/padded in some way.

> Btw...
> Since the filesystem is mostly read-only, I was also thinking about
> reducing the size of the log section. The planned filesystem will be
> 5.5T or bigger, which results in 2G log section per default. I thought
> about limiting it to 128M or 64M. There will be no concurrent write
> activity on the fs. Together with the zero-alignment, what do you think
> about that?

That'll be fine - the maximum log size until a couple of years ago
was 128MB, and I still do most of my benchmarking with a log that
size....

Cheers,

Dave.
-- 
Dave Chinner
david@fromorbit.com

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs