sar -v report and HFS inodes in system.

[CAS_2]

Hi.

I read that ninode (HP-UX tunable kernel parameter) refers to HFS i-nodes, not to VxFS i-nodes.

In my box there is only one HFS file system (/stand). It contains 66 i-nodes.

But 'sar -v' reports there are 1188 entries used in ninode table. 1188 entries from 11332. 11332 is the current value of ninode parameter.

How can this possible ?

How are there more used HFS i-node entries when there are not enough HFS i-nodes in the only HFS file system ?

[Jaime Bolanos Rojas.]

Hola CAS!

Where did you get that information CAS!

ninode refers to the maximum number of open inodes that can be in memory.

Take a look at this link for more information:

http://docs.hp.com/en/TKP-90202/re58.html

Regards,

Jaime.

[CAS_2]

ninode(5) man pages on HP-UX 11.23 v2
ninode(5)
Tunable Kernel Parameters

NAME
ninode - maximum number of HFS file system open inodes that can be in
memory

VALUES
Failsafe
476

Default
The default value is computed at runtime and depends on the amount of
physical memory on the system. For small memory systems (less than
1GB), the default is 4880. For systems with more than 1GB of memory,
the default is 8192, or 8K.

Allowed values
The minimum value allowed is 14. The maximum value allowed is memory
limited.

Specify a positive integer value.

DESCRIPTION
The ninode tunable defines the number of slots in the HFS inode table.
This number limits the number of open inodes that can be in memory for
HFS file systems at any given time. The inode table is used as a
cache memory. For performance reasons the most recent ninode (number
of) open inodes are kept in main memory. The table is hashed.

Each unique open file has an open inode associated with it.
Therefore, the larger the number of unique open files, the larger
ninode should be.

Who Is Expected to Change This Tunable?
This tunable is only relevant on systems where there are HFS file
systems mounted. Note that the use of HFS file systems is being
deprecated in HP-UX.

[Jaime Bolanos Rojas.]

Hola de nuevo CAS!

OK, I see what's going on right now, you are running 11.23,

And you decided to used /stand as a HFS file system, then the information that you showme in the ninode(5) is right, but it does not mean that ninode went away, they created vx_ninode http://docs.hp.com/en/B2355-60105/vx_ninode.5.html

which most likely is going to show with the same sar -v command, becuase it's referring to the same thing, but different file systems formats.

Hope that answers your question,

Regards,

Jaime.

[CAS_2]

No, my box is running 11.11, but the fact that ninode is referring to HFS i-nodes instead of all-system i-nodes is also found in several threads in these forums. For example, in thread

http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=1040911

Bill Hassell wrote

> ninode is more than 10x too large. Change it to 2048 and forget it. This parameter is primarily for in-core inodes from the /stand filesystem.

My question is due that Oracle pre-requisites suggest to set ninode tunable parameter to

8 * nproc + 2048

But if ninode refers only to HFS, it is unnecessary.

Then I checked ninode current value by means of sar in order to confirm this and ignore formulae for ninode.

[Jaime Bolanos Rojas.]

Hi CAS!

Yes you are right and Bill was right, there is a reason why sar -v displays 1188 entries in your ninode table, even thought you are just using 66 i-nodes in /stand.

That is explained very well by this HP doc located in the ITRC site.

DocId: KBRC00002842

Regards,

Jaime.

[Dave Olker]

Hi CAS,

Most likely the Oracle documentation recommends to size ninode at that value because either:

1) The documentation was written back before VxFS inodes had their own tunable

2) Oracle knows that ninode is used to size other kernel tunable parameters

It could be either one, but the second point is still true - ninode is used in many formulas to size other kernel tunable values. So while Oracle likely doesn't actually need lots of HFS inodes, they may require that other tunables are set higher than their default value and they're using ninode as a way to indirectly size them.

You might want to contact Oracle and ask them if they have updated their recommendations to outline exactly which kernel tunables they recommend increasing and why.

Regards,

Dave