vmstat odd output.

[Todd McDaniel_1]

I have a question about vmstat.

I have a DBA asking me about the CPU portion: usr and sys... It seems they always show 0 0 100 on the CPU portion except for the first line which has very unusual numbers under CPU...

Can anyone help me figure this out?
us sy id
80 25 226066174050304
0 0 100
0 0 100
0 0 100

Below is the full line sorry for the line wrap...


$ vmstat 5 4
procs memory page faults cpu
r b w avm free re at pi po fr de sr in sy cs us sy id
6 22 0 643445 10709411 225 69 0 0 0 0 0 19826 47796 6774 80 25 226066174050304
6 22 0 643445 10715199 247 26 0 0 0 0 0 36357 61601 15388 0 0 100
13 19 0 623823 10714653 238 10 0 0 0 0 0 34456 58239 13746 0 0 100
13 19 0 623823 10724254 352 54 0 0 0 0 0 32908 54598 12395 0 0 100

[Steven E. Protter]

Its hard to figure out your output.

I'm attaching some sar scripts that might get you better performance data.

They are background, production, originated from HP.

SEP

[Todd McDaniel_1]

Sorry for the output I tried to cut out the CPU portion to make it readable...


My question is this: With the command, "Vmstat 5 5", why does the IDLE time show 100% and 0 for the user and sys columns... I ran vmstat 5 100 and every time it shows:


usr sys id
0 0 1000


Every time except for the first time which seems corrupted...And this is during hte day with some load on the system so it should not always show idle 100 every 5 seconds???

I must say this is on my Superdome 11.11... All my other boxes show some variation in these last 3 colums... any thoughts no matter how weird are welcome

[Francis_12 · Accepted solution]

Hello,

What is your vmstat patching level ?

PHCO_27578 is the latest one for 11.0, PHCO_25979 is the latest for 11.11.

I can read from patch texts that both of them are solving CPU wrong display outputs.

Hope this helps, Bye.

Francis DERDEYN - HP-UX ASCE.

[Francis_12]

Hello back again,

I just saw your addtionnal mail saying that this is a Superdome system. So, you are totally matching the 'known' problem solved by PHCO_25979 :


vmstat(1) is showing wrong value for average CPU idle time
spent by all the active processors. The average cpu time
spent in each state was calculated by storing the sum of
cpu time spent in each state by all the active processors
in a variable and dividing it with number of active
processors. There was a overflow occurring while assigning
the sum of cpu times spent by all the active processors.
This problem can occur only on multi-processor systems, up
for long time.

Can you verify if you have that patch installed. If not, install it and check back again. The patch does not need any reboot.

Bye.

Francis DERDEYN - HP-UX ASCE.

[Michael Steele_2]

Here are 11.00 patches:

s700_800 11.00 vmstat(1) cumulative patch PHCO_27578 PHCO_27578
s700_800 11.00 pstat patch, long command line storage PHKL_26008

Here is the 11.11 patch:

s700_800 11.11 vmstat(1) cumulative patch PHCO_25979 PHCO_25979

[Todd McDaniel_1]

I think I need to make that patch... I have notified my Platform specialist regarding this issue...

It describes my exact problem.

[Todd McDaniel_1]

Patch worked like a charm. Loaded it onlinie and vmstat worked right after that.