[SOLVED] Changing XFS sunit and swidth parameters on RAID 5

Zeratul021 · 11-05-2009, 09:55 AM

Hi all,

I would like to ask something regarding XFS and RAID 5:

I've built raid 5 from 4 1TB devices using 256KB chunk size.
I wanted to tweak XFS on it a little so I formatted it with

Code:

mkfs.xfs -f -L DATA -l internal,size=128m -d agcount=30,su=256k,sw=3 /dev/md0

Problem is that output shows:

Code:

eta-data=/dev/md0               isize=256    agcount=30, agsize=24419072 blks
         =                       sectsz=4096  attr=2
data     =                       bsize=4096   blocks=732571776, imaxpct=5
         =                       sunit=64     swidth=192 blks
naming   =version 2              bsize=4096   ascii-ci=0
log      =internal log           bsize=4096   blocks=32768, version=2
         =                       sectsz=4096  sunit=1 blks, lazy-count=0
realtime =none                   extsz=786432 blocks=0, rtextents=0

That's not exactly sunit that I wanted (should be 512 and swidth 1536).
Even if I try to mount it with these options set to proper numbers, xfs_info still provides same output.

Any ideas or suggestions why this happens?

Regards,

Marek

Zeratul021 · 11-10-2009, 03:01 PM

Solved:

In reference to http://oss.sgi.com/archives/xfs/2003-04/msg00145.html

where it is stated that:

Quote:

data = bsize=4096 blocks=84286944,
imaxpct=25
= sunit=16 swidth=48 blks,
unwritten=0

since it apparently displays the "sunit" and "swidth" values in terms of
XFS blocks, which are 4096 bytes each....

That means that my 64*4096 is my 256k and swidth 192*4096/512 is my 1536.

Anyway, formatting without parameters resulted in the same numbers because xfs' built-in optimization for software raids.