mkfx with option -o ninode=100 has no effect

[Mauro Cimino_2]

Hi, I use the command mkfs to create a file system and let on it a low space for the inode ( I havo to create on this file system only some large file). I use the option "-o ninode=100", baut when I mount the file system I see that this let a big space for inode. And if I do "mkfs -m " to see the used option I doesn't show me the ninode parameter. The O.S. is 11.23. Anybody knows something on this problem? Mauro

[OFC_EDM]

How many inodes are on the mount point?

bdf -i will show you

[OFC_EDM]

Which shell are you using?

I assume you ran the mkfs as root.

Search your history file for the mkfs command you used and find out what you specified for inode.

Assuming you're using bash and roots homedir is /
grep -i "mkfs" /.bash_history

For SH / KSH
grep -i "mkfs" /.sh_history

Cheers

[YAQUB_1]

If U taking about ninode kernal parameter....
the default value of ninode is defined in terms of nproc, maxusers and npty.

[Mauro Cimino_2]

the comman I use is :
mkfs -o ninode=10000 -o largefiles /dev/vg_fc_csost_2/lv_csost_2
and the shell is /sbin/sh.

Note, I have another system , with 11.23, and on this one the command mkfs works properly, with the width I want. On the first, instead, the width is always then same, indipendent of parameter given on the command line.

Perhaps some kernel parameter has some problem?

[Rasheed Tamton]

Hi,

mkfs -F vxfs -m /dev/vg00/lvol6
mkfs -F hfs -m /dev/vg00/lvol1

What is you filesystem type. Is it hfs or vxfs.

What commands you used, when you created the filesystem.

Rgds.,
Rasheed Tamton.

[Mauro Cimino_2]

The problem is that on the 280 Gb logical volume, I don't want to dissipate 18 Gbyte, because I need only some big database file on disk.

[Mauro Cimino_2]

The file system type is vxfs.

[YAQUB_1]

can u send kctune output....

[Mauro Cimino_2]

Attachment is the output of kctune command.

[Rasheed Tamton]

Is this filesystem largefiles enabled.

:> mkfs -F vxfs -m /dev/vg00/lvol6
mkfs -F vxfs -o ninode=unlimited,bsize=8192,version=5,inosize=256,logsize=2048,largefiles /dev/vg00/lvol6 1048576

Just post the command you used.

[Mauro Cimino_2]

mkfs -o ninode=10000 -o largefiles /dev/vg_fc_csost_2/lv_csost_2
this is the command I used.

and the output is:

mkfs: /etc/default/fs is used for determining the file system type
version 6 layout
280231936 sectors, 280231936 blocks of size 1024, log size 16384 blocks
largefiles supported.

As you can see, it is like i dont'use the option -o ninode=10000.

When I do the same command on another system, I get:
mkfs: /etc/default/fs is used for determining the file system type
version 5 layout
419364864 sectors, 419364864 blocks of size 1024, log size 16384 blocks
100 inodes, largefiles supported
419364864 data blocks, 419242832 free data blocks
12798 allocation units of 32768 blocks, 32768 data blocks.
I see that it use the parameter.

[Venkatesh BL]

What is the content of /etc/default/fs file on both the systems?

[Rasheed Tamton]

Hi,

The apparent difference with both the systems is the second one uses version 5 and the first one is using version 6 layout.

What is the differnce with both the systems.
Please see:

man mkfs_hfs

[Rasheed Tamton]

Also see
man mkfs_vxfs

[Mauro Cimino_2]

On this system only teh man of the mkfa_vxfs and mkfs_hfs exists. not the executable file.
find / -name mkfs_vxfs
find / -name mkfs_hfs

No file has been founded.

[Rasheed Tamton]

There are no commands like those. It is related to mkfs.

> The apparent difference with both the systems is the second one uses version 5 and the first one is using version 6 layout.

>What is the differnce with both the systems.

Are both the systems same? Why on one it is
version 5
and
on the other
version 6

[Mauro Cimino_2]

On the system where mkfs is working properly:

root@carnot:/>what /usr/sbin/mkfs
/usr/sbin/mkfs:
$Revision: B11.23_LR
Fri Aug 29 21:16:05 PDT 2003 $
root@carnot:/>

On the other system:
root@mrh-ora2:/>what /usr/sbin/mkfs
/usr/sbin/mkfs:
$Revision: @(#) wrappers R11.23_BL2005_0504_3 PATCH_11.23 PHCO_32611
root@mrh-ora2:/>

[Rasheed Tamton]

I think you have to install the patch PHCO_32611 on the other system. It is a fs_wrapper(5) patch and will update many files in /sbin and /usr/sbin related to file system commands including mkfs

If the HPUX VxFS4.1 feature bundles (FEATURE11i, EnableVxFS and Base-VXFS) are not installed, this product update will have no impact on your system.

http://www11.itrc.hp.com/service/patch/patchDetail.do?BC=main|search|patchDetail{PHCO_32611,{hpux:11.23,}}|&patchid=PHCO_32611&sel={hpux:11.23,}

Rgds.
Rasheed Tamton

[Mauro Cimino_2]

Ramsheed,

the patch you said is installed on the system that is not working. Mauro

[Rasheed Tamton]

So, what I come to know is that your working system is using vxfs version 5 and the system not working is using vxfs version 6. If your sytem has version 5 also installed the below command might work:

Before you change, MAKE SURE that you have the BACKUP of the data resides on the LV. I would also recommend to unmount /mountpoint

Here I am using /dev/vg00/lvol9 for test:

/opt/VRTS/bin/mkfs -o version=5 /dev/vg00/lvol9

or
/sbin/mkfs -o version=5 /dev/vg00/lvol9

and then run mkfs with ninode option as you tried before.

or go to
-------------- reminder backup of the data ----
cd /sbin/fs/vxfs
and change the version to 4 or 5

./mkfs -o version=4 /dev/vg00/lvol9
./mkfs -o ninode=1000 /dev/vg00/lvol9
./mkfs -m /dev/vg00/lvol9

you can also change the version to 5 as below:
./mkfs -o version=5 /dev/vg00/lvol9
./mkfs -o ninode=1000 /dev/vg00/lvol9
./mkfs -m /dev/vg00/lvol9

/sbin/fs/vxfs:> ./mkfs -m /dev/vg00/lvol9
mkfs -F vxfs -o ninode=1000,bsize=1024,version=5,inosize=256,logsize=16384,nolargefiles /dev/vg00/lvol9 2129920

mount the LV.

bdf -i /mountpoint

Regards,
Rasheed Tamton.

[Mauro Cimino_2]

Yes, I succed to create a file system with only 100000 inode, Ramsheed.

I changed in /sbin/fs/vxfs

and :
./mkfs -o versione=5 -o ninode=100000 ....

and so on.

Your solution is working. Thank You , Ramsheed

[Rasheed Tamton]

Hi Mauro,

Good to know it is solved. I am Rasheed not Ramsheed (without m). Hope you would assign points.

Regards,
Rasheed Tamton.