increase inode value in VXFS filesystem

[Jagadesh_2]

Please let me know how to increase the inode count in VXFS file system (HPUX 11.11). I have very low inode count in that volume. The same volume is used for running SAP application.


Thanks
S.Jagadesh

[Robert-Jan Goossens]

Hi,

Check this doc.

http://www8.itrc.hp.com/service/cki/docDisplay.do?docLocale=en_US&docId=200000064129598

Document description: Inode Limitation on VxFS w/o Large File Support
Document id: UFSKBRC00010611

Regards,
Robert-Jan

[Govinda_1 · Accepted solution]

hi,

http://docs.hp.com/en/B2355-60103/vx_ninode.5.html

see the man pages attached. u can tune vx_ninode .

hope this helps,

Regards,
Govinda G.

[Eric Antunes]

Hi,

I think a good start, if you have a lower value, will be 2048. To understand the need to increase, check possible overflows with:

#sar -v 5 1000

Best Regards,

Eric Antunes

[Jagadesh_2]

bdf -i displays

Is it possible to increase the inode count in Logical volume dynamically?

How to go about in rectifying the same.

[Jagadesh_2]

Hi Eric,

sar -v 5 1000 display
the o/p is attached along with this

[Eric Antunes]

Hi,

I think it is useless to increase inode since your 7360 value is far from being reached.

Your nproc and nfile parameters seems also a bit exagerated but you need to collect more than a minute of data to be more certain about that. Try to run:

#sar -v 5 100 > sar_5_100.log

After this, analyse that log file to see if you have any overflow situation (any "ov" column different than 0)...

Best Regards,

Eric Antunes

[Bill Hassell]

You can ignore the bdf -i listing as well as sar -v as far as inodes. Unlike HFS filesystems where the inode count is fixed by newfs/mkfs, VxFS filesystems build inodes as required. Therefore, bdf -i is just a sampl number and will increase automatically. It won't fill up until all disk space in the filesystem is also full.

sar -v reports an inod value that has nothing to do with individual inodes on a filesystem. This is an in-core cache of HFS inodes, used to speed up re-opens of the same files. It gets a bit complicated because this cache cannot be measured for reusable entries. sar -v (after a few days) will show inode = 99%-100% full, yet it the cache continues to function. If you leave ninode (the kernel parameter for this metric) fixed at 4000, you'll be fine. You can test the automatic inode creation feature for VxFS by creating a small lvol and creating dozens of directories and thousands of files. Run bdf -i to show the progress and you'll see the ifree value is fairly constant while the iused value increments without bounds.

[Jagadesh_2]

Thanks Bill,

Based on your reply. I had came to a conclusion that there is no problem for inode count in VXFS file system.