kmeminfo 11.00 uses too much cpu utilization

[azztigin]

hi,

What is kmeminfo 11.00 and why does it take up so much cpu utilization?

Here is the result of my top command:

TTY PID USERNAME PRI NI SIZE RES STATE TIME %WCPU %CPU COMMAND
? 27906 root 233 20 12876K 12920K run 12820:11 48.73 48.65 kmeminfo.11.00
? 15896 root 234 20 12876K 12616K run 3523:34 48.38 48.30 kmeminfo.11.00
? 29 root 152 20 0K 0K run 13:37 1.17 1.17 vxfsd
? 13180 root 152 20 14396K 51972K run 14:11 1.05 1.05 java
? 363 root 154 20 32K 40K sleep 25:43 0.13 0.13 syncer
pts/ta 7205 root 168 20 2680K 464K run 0:00 0.12 0.12 top
? 6364 root 168 20 2712K 472K sleep 38:33 0.12 0.12 top
? 3 root 128 20 0K 0K sleep 15:08 0.06 0.06 statdaemon
? 6947 root 154 20 204K 296K sleep 0:00 0.03 0.03 registrar
? 1644 root 154 20 112K 140K sleep 1:47 0.03 0.03 cron
? 0 root 128 20 0K 0K sleep 0:15 0.02 0.02 swapper
? 1 root 168 20 460K 220K sleep 1:29 0.02 0.02 init
? 2 root 128 20 0K 0K sleep 0:56 0.02 0.02 vhand

[Olivier Decorse · Accepted solution]

Hi,
kmeminfo.11 is a tool that help you to see what memory is available, used and how is is used : static, dynamic, for bufferCache, etc.
Just run /usr/contrib/Q4/bin/kmeminfo.11.00 and you will see something like this :
libp4 (4.14): Opening /stand/vmunix /dev/kmem

Processing pfdat table (251379 entries)...

----------------------------------------------------------------------
Physical memory usage summary (in pages):

Physmem = 262144 Available physical memory:
Freemem = 85973 Free physical memory
Used = 176171 Used physical memory:
System = 149062 by kernel:
Static = 10765 for text and static data
Dynamic = 41239 for dynamic data
Bufcache = 97058 for file-system buffer cache
User = 30699 by user processes
Uarea = 1520 for thread uareas
Disowned = 512 disowned pages
[...]

It is normaly that kmeminfo.11 use a some cpu, BUT, it has to stop, and it is not normaly to see it always on top (2 times !!) !!
Somebody has certainly run it, and it can't stop.
I suggest you to run it manually, and see any error or any freeze display.
Let us see the result.

Olivier.

[R. Sri Ram Kishore_1]

Hi,

Reg. what kmeminfo is, check out these threads:
http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=264147
http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=543227

HTH.

Regards,
Sri Ram

[azztigin]

here is the result:

#/usr/contrib/Q4/bin/kmeminfo.11.00
libp4 (4.14): Opening /stand/vmunix /dev/kmem

Processing pfdat table (250441 entries)...

----------------------------------------------------------------------
Physical memory usage summary (in pages):

Physmem = 262144 Available physical memory:
Freemem = 74574 Free physical memory
Used = 187570 Used physical memory:
System = 131244 by kernel:
Static = 11703 for text and static data
Dynamic = 54005 for dynamic data
Bufcache = 65536 for file-system buffer cache
User = 59260 by user processes
Uarea = 1448 for thread uareas
Disowned = 512 disowned pages

----------------------------------------------------------------------
Dynamic memory usage summary (in pages):

Dynamic = 54005 Kernel dynamic data (sysmap):
MALLOC = 9038 memory buckets:
bucket[ 5] = 355 size 32 bytes
bucket[ 6] = 32 size 64 bytes
bucket[ 7] = 843 size 128 bytes
bucket[ 8] = 270 size 256 bytes
bucket[ 9] = 559 size 512 bytes
bucket[10] = 668 size 1024 bytes
bucket[11] = 4093 size 2048 bytes
bucket[12] = 44 size 4096 bytes
bucket[13] = 112 size 2 pages
bucket[14] = 31 size from 3 to 4 pages
bucket[15] = 11 size from 5 to 8 pages
bucket[16] = 123 size from 9 to 16 pages
bucket[17] = 132 size from 17 to 32 pages
bucket[18] = 247 size from 33 to 64 pages
bucket[19] = 74 size from 65 to 128 pages
bucket[20] = 1444 size from 129 to 256 pages
Kmalloc = 44929 k[m]alloc()
Eqalloc = 24 eqalloc()
Reserved = 14 Reserved pools

[Olivier Decorse]

the "Reserved pools" is the also the last line of my result.
Does it correctly finished ?
can you see the prompt after ?

I think you have to kill the 2 other kmeminfo in memory and see if it works better.
If not, you can open a call : there is certainly a patch that update kmeminfo.
What is your OS version ?

And finally, as mentioned in a ohter thread :
You can get tusc and see where it's hanging with :
tusc -vfp "kmeminfo"

Olivier.

PS : don't forget to assign points, if any response help you.

[azztigin]

yes, it did go back to prompt. The OS version is 11.00.

[Olivier Decorse]

ok, so there is no hang problem with your kmeminfo.
You can kill the 2 in memory, and you will return to a normal cpu utilisation.

The last question is : why are those 2 kmeminfo in memory ?
1) have you run it ?
2) can anyone else have run it : other system administrator ?
3) are there run from startup of the server, to diagnose any problem ?

Good luck. Olivier.

PS : thank you very much for your points you have assigned !