# machinfo CPU info: 12 Intel(R) Itanium(R) Processor 9350s (1.73 GHz, 6 MB) 2.39 GT/s QPI, CPU version E0
Memory: 28026 MB (27.37 GB)
Firmware info: Firmware revision: 04.30.00 FP SWA driver revision: 1.18 IPMI is supported on this system. BMC firmware revision: 4.1e
Platform info: Model: "ia64 hp server Integrity Virtual Machine" Machine ID number: 2ba29fb2-b854-11e0-8718-b499ba645408 Machine serial number: VM01130004
OS info: Nodename: --------- Release: HP-UX B.11.31 Version: U (unlimited-user license) Machine: ia64 ID Number: --------- vmunix _release_version: @(#) $Revision: vmunix: B.11.31_LR FLAVOR=perf
Reviewing the current kernel settings I've noticed an particular behavior on ninode value. At present machine have configured the 60096 value on ninode parameter:
#kctune -d ninode
Tunable Value Expression
Description
ninode 60096 60096
Maximum number of HFS file system open inodes that can be in memory.
By other way, when I run kcusage the following values are displayed:
#kcusage ninode
Tunable Usage / Setting
=============================================
ninode 2346 / 508439
As you can see the setting value is 579903 instead 60096!
So, when I execute sar -v the fllowing values are displayed:
# sar -v 1 10
HP-UX produ2 B.11.31 U ia64 11/22/11
14:33:04 text-sz ov proc-sz ov inod-sz ov file-sz ov
At this point the inod-sz is the same that kcusage value (508.439). I'm really confused by this situation. How I can controll the real value of ninode at kctune session?
Also I can see a highest value of file-sz (2.147.483.647) How I can set a discrete value?
We are trying to get free memory. Please find attached a file with kctune dump.
ninode has been historically messed up as a setting sine before version 10. Always override ninode with a fixed value of 4096 or 8192. The formula is worthless and creates a massively large table in the kernel that will never be used.
