kcusage report filecache_max value bigger than tunable

[unixdaddy]

I'm wondering why on my 11.31 (march 2011) rx2800 i2 with 47GB of memory and a filecache max of 5% i see consistently higher values in kcusage? both running oracle 11gr2.

kctune filecache_max
Tunable             Value  Expression  Changes
filecache_max  2445724467  5%          Imm (auto disabled)

kcusage filecache_max
Tunable                 Usage / Setting
=============================================
filecache_max      2971787264 / 2445724467

kcusage -d filecache_max
Tunable:        filecache_max
Setting:        2445724467
Time                           Usage      %
=============================================
Sun 11/27/11 18:00 GMT    2436235264   99.6
Sun 11/27/11 19:00 GMT    2436235264   99.6
Sun 11/27/11 20:00 GMT    2436239360   99.6
Sun 11/27/11 21:00 GMT    2436235264   99.6
Sun 11/27/11 22:00 GMT    2436235264   99.6
Sun 11/27/11 23:00 GMT    2436235264   99.6
Mon 11/28/11 00:00 GMT    2436235264   99.6
Mon 11/28/11 01:00 GMT    2436308992   99.6
Mon 11/28/11 02:00 GMT    2436259840   99.6
Mon 11/28/11 03:00 GMT    2436235264   99.6
Mon 11/28/11 04:00 GMT    2436235264   99.6
Mon 11/28/11 05:00 GMT    2436235264   99.6
Mon 11/28/11 06:00 GMT    2436239360   99.6
Mon 11/28/11 07:00 GMT    2436235264   99.6
Mon 11/28/11 08:00 GMT    2436235264   99.6
Mon 11/28/11 09:00 GMT    2436308992   99.6
Mon 11/28/11 10:00 GMT    2436235264   99.6
Mon 11/28/11 11:00 GMT    2436235264   99.6
Mon 11/28/11 12:00 GMT    2972860416  121.6
Mon 11/28/11 13:00 GMT    2972860416  121.6
Mon 11/28/11 14:00 GMT    2972860416  121.6
Mon 11/28/11 15:00 GMT    2972860416  121.6
Mon 11/28/11 16:00 GMT    2972860416  121.6
Mon 11/28/11 17:00 GMT    2972860416  121.6

I'm thinking maybe to up the max to 8%

[Duncan Edmonstone]

The only situation I can see where this would make sense is where filecache_max has been set to a hard value, rather than a percentage, and then a memory OLD event has occurred to reduce the total amount of system memory.

But that ceratinly doesn't look like the case on your system, as we see the value actually go _up_, so I'm at a bit of a loss to explain that... some things to check/look at...

- Any recent changes of kernel parameters (kclog will tell you)

- I'd make sure you have the latest VM patch (PHKL_41969) and associated dependencies on the system

- Have a look at the log file for kcmond just to be sure it isn't chucking errors (/etc/opt/resmon/log/api.log)

- 47GB seems an odd amount of memory - is it not 47.9GB (i.e. really 48GB) - what does machinfo say?

[unixdaddy]

Hi Duncan,

we worked together on a retail business in north west a couple of years back ;-)

well nothing has been done to these boxes since they were built - no kernel changes, the patch isn't installed (so i'll look to do that) and yes you are right about the memory it is 47.9Gb i'll check out the api.log.