maximum block size for vxfs filesystem

[Kirk Reindl]

HPN4000
HP-UX 11.i
Extent based striping
Oracle 9.0.1

I have been asked by my Oracle DBA to create vxfs filesystems using 16K block sizing. My volume group is set up with the default 4MB Physical Extent Size.

When I try to make the filesystem this is the response I get:

#newfs -F vxfs -b 16384 /dev/vgdss1/rpdss01
vxfs mkfs: bsize must be a power of 2 >= 1024 and <= 8192

Can anyone please tell me how I can create a filesystem with 16K block size? Is this possible with vxfs? Do I need to change a kernel parameter to make this work?

Any answers or documention clarifying my question would be appreciated.

Thanks

Kirk

[S.K. Chan]

In vxfs FS the largest block size you use is 8K. Nothing you can do to increase it. Typically (in 11.x) when a vxfs FS is created even without the "-b" option it'll default to 8K block size. If you run ..

# fstyp -v /dev/vgXX/rlvolX

on any of your existing vxfs lvol and look for
the value of "f_bsize" you'll see it is set to 8192.

[Printaporn_1]

Hi,

Just make it 8k as maximum , Oracle Block is multiple of OS block.

[Roger Baptiste]

kirk,

The maximum value you can specify is 8192. from the man page of mkfs_vxfs :

***
bsize=n n is the block size for files on the file system and represents the smallest amount of disk space that will be allocated to a file.
n must be a power of 2 selected from the range 1024 to 8192. The default is 1024
***

newfs -F vxfs calls mkfs_vxfs to do the job. and the -b you specify in newfs should be between 1024 to 8192 (powers fo 2).


8KB block size is good enough. It is a close multiple of the oracle Block size .


SK,
Please note that the default for newfs is not 8192! It is 1024 (1Kb) only! . The value which denotes this is f_frsize (and not f_bsize). So, it is always useful to give -b 8192 option while creating a oracle filesystem.


HTH
raj

[Wodisch]

Hello,

we had that kind of discussion recently over there:
http://forums.itrc.hp.com/cm/QuestionAnswer/1,,0x6dc0c8ecad09d6118ff40090279cd0f9,00.html

HTH,
Wodisch

[A. Clay Stephenson]

The good news is that vxfs filesystems are not really sensitive to block sizes unlike hfs filesystems. Vsfx filesystems tend to write in 64K chunks no matter what the block/extent size is. The main concern is making your LVOL stripe size 64K to match the preferred 64K i/o chunks of vxfs (if you stripe, that is).