inod question

[Juan Gonzalez_2]

I have a server that has 30,000 inodes max configure and a file system that uses 5,000 inodes. I split the bc of that filesystem and mount that filesystem to another server that has 4,000 inodes configured will this cause me a problem backing up the filesystem because I do not have enough inodes.

[James R. Ferguson]

Hi:

If the filesystem is a VxFS one then inode allocation is dynamic. If you need more inodes, expand the filesystem.

Regards!

...JRF...

[John Palmer]

If you're worried about the kernel parameter 'ninode', don't.

'ninode' only defines the size of a cache of recently used inodes that the kernel uses. It doesn't restrict the number that the kernel can handle.

Regards,
John

[A. Clay Stephenson · Accepted solution]

Note that ninode only applies to hfs filesystems. I'll bet that the only hfs filesystem on your box is /stand. If that is the case, you can drastically reduce niniode to a few hundred and save resources. Typically, the maximum number of inodes will be in play in building kernels in /stand and for that a few hunderd will more than suffice.

[Juan Gonzalez_2]

John then what kernel parameter can tell me how many inodes my server can handle and my file system is vxfs.

[A. Clay Stephenson]

There is no such parameter for vxfs inodes; there is nothing to examine. The number of inodes needed for vxfs is increased as needed automatically. There is an obsolete tunable called vx_ninode but it is completely ignored. Again, if your only hfs filesystem is /stand then reduce niniode to something around 500 and forget about it.

[James R. Ferguson]

Hi (again) Juan:

As we have noted, if the filesystem is VxFS then there is no inode limit.

Are you by any chance thinking of the maximum number of sub-directories of a directory?

If so, there is a tunable 'vx_maxlink' which allows this limit to reach 64K-1 (PHKL_27648).

Regards!

...JRF...

[Juan Gonzalez_2]

Thanks everyone for your information and hope you all have a great Halloween.