How do I change the nbpi (inode size) value

that would be the npbi option of nas_fs

however since the default is 8k and the Celerra block size is also 8k (no partials) this usually doesnt make sense unless you have lots of empty files or links/symlinks

I think what you are rather hitting is the ufs.inodelimit param

ufs inodelimit
257949696 ¿ 4294967295
Default: 257949696 (0xf6000000)
Specifies the maximum number of inodes for a new file system or a file system extension.

In order to save space by default we only create inodes for the first 2TB worth of file system.
If you change that param and then create a 6TB fs you get more inodes
Also if you extend it (I believe)

Rainer, the application that uses this space creates small txt files that it gathers from the web. Our standard deviation on the file sizes range a lot - largest are 250K, smallest are a handful of bytes.

Once the upper limit is increased and the datamover rebooted is there any way to increase the number of inodes without destroying/recreating the volume?

Luckily we have an active month and a previous month.  Since this is the first go-round with the Celerra I had additional space to create new volumes after adjusting the ufs.inodelimit to 1.5 billion.  I did do it with Celerra manger but I took the failover reboots just to be safe.   I lowered my inode size to 4k just to be sure we could make it to the end of the month.

Command I ran:

nas_fs -name d02 -create size=6400000M pool=Pool-d02 -o slice=y -o nbpi=4096

Here is the inode math.  Please correct me if I am wrong:

Get from TB to KB:

volume size 6.2TB * 1024 * 1024 *1024 = 6657199308KB

If I have a 8k inode I would have 6657199308.8/8 = 832149913 total inodes

If I have a 4k inode I would have 6657199308.8/4 = 1664299827 total inodes

Once the application is done we will take another look but I have a feeling we will be moving back to the 8k inode size.  Thanks for all your help!

Hi Mike,

just curious - what did you decide to do ?

Rainer

Hi Mike,

looks fine

with 8k nbpi you should be getting 128M (128*1024*1024) inodes per TB

and for 4k nbpi twice of that

you're welcome

Rainer