CPU load high, top processes very low?

Thinking · 04-20-2005, 04:43 AM

hiho@ll

i programmed my own reporting client/server which reports host load

it ran last night, very well, but i don't understand this:

the logging intervall is every 30 seconds
i log the cpu load using /proc/stat
and i log the top 3 processes using top

my problem now is
i got some very high cpu load spikes over a half an hour
but the top processes were as low as ever!?

is it possible, that the /proc/stat cpu load is high but, top reports nearly sleeping processes?
maybe /proc/stat is a relative

for top i'm using the command
top -n 1 -b

what also could be a problem is that top influences it's own statistics
which means
if i run top -n 1 -b and the cpu is idle, the top process will consum high cpu time

which could also influence my measures

anybody knows an alternative to top?
can i read in /proc which processes consume most cpu?

thx@ll

Geronimo · 04-20-2005, 05:11 AM

try runnig `top`... And have a look at the top entries - i.e. the ones that use most CPU time. hit 'q' to quit!

Hope it helped

Greets
G.

Thinking · 04-20-2005, 09:05 AM

well

as i posted i'm using the "top" command

and my question is:
is it possible, that cpu load is very high, but "top" reports just 3.8% cpu for the process "top" and every other process is nearly 0.0%?

thx

Gamezace · 05-24-2005, 12:49 PM

I just had a very similiar problem - and so I ran gkrellm juxtaposed to a terminal running 'top' and found interesting results.

gkrellm was reporting my CPU load abot 50% constantly, whereas top was no where about 18% most of the time...not really sure what is causing that.

Gamezace · 06-03-2005, 01:37 PM

Well, I solved the problem on my machine. As it turns out, the problem was transparency. I had all my terminals set to transparent backgrounds - and I guess this old company machine doesn't have enough power to handle that.

Maybe this will help you,

sahithi · 03-15-2007, 01:48 PM

top shows a process taking 98% of CPU time

but

cat /proc/stat gives the idle jiffy time a constant value which does not decrease..

is something wrong with /proc/stat or is top showing it wrong?

what command exactly shows the cpu stats ?

studioj · 03-15-2007, 09:45 PM

well just basically you got to dig.
all the stats in /poc got limitations thats for sure, or they would have to spend lots of time doing /proc junk instead of writing system code.

its very confusing when you say tranperance fixed BUT X was not showing up as the culprit.
see X has the video card memory and what not to deal with
perhaps it would help also to turn down RT tick frequencies ? do you interupt frequencies turnrd way up ?

i killed a P4 like that once faster than you can say, "why does my system keep freezing up"

syg00 · 03-15-2007, 10:23 PM

Edit: D'oh - didn't realize I responded to such an old post.
Deleted.

edenCC · 03-16-2007, 04:03 AM

top is not so good for your detection, and using ps will be much better.

example:
top 3 cpu consumption
ps -auxf | sort -nr -k 3 | head -3
top 3 mem consumption
ps -auxf | sort -nr -k 4 | head -3

Crito · 03-16-2007, 04:49 AM

Subtract top's cpu idle time (%id) from 100% and you'll see it's the same number gkrellm is reporting. I think they're both getting that info from vmstat anyway.

Code:

[climber@lode ~]$ top

top - 05:33:19 up 16 days, 13:48,  2 users,  load average: 0.09, 0.20, 0.18
Tasks: 141 total,   2 running, 139 sleeping,   0 stopped,   0 zombie
Cpu(s):  0.0%us,  0.0%sy,  0.0%ni, 99.7%id,  0.0%wa,  0.3%hi,  0.0%si,  0.0%st
Mem:    954560k total,   902380k used,    52180k free,   196300k buffers
Swap:  4192924k total,    38984k used,  4153940k free,   355816k cached

  PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+  COMMAND            
27584 root      16   0 12708 1076  792 S  0.3  0.1   0:00.27 top                
    1 root      15   0 10304  568  540 S  0.0  0.1   0:00.73 init               
    2 root      RT   0     0    0    0 S  0.0  0.0   0:00.00 migration/0        
    3 root      34  19     0    0    0 S  0.0  0.0   0:10.10 ksoftirqd/0        
    4 root      RT   0     0    0    0 S  0.0  0.0   0:00.00 watchdog/0         
    5 root      10  -5     0    0    0 S  0.0  0.0   0:18.86 events/0           
    6 root      12  -5     0    0    0 S  0.0  0.0   0:00.00 khelper            
    7 root      13  -5     0    0    0 S  0.0  0.0   0:00.00 kthread            
    9 root      10  -5     0    0    0 S  0.0  0.0   0:00.00 xenwatch           
   10 root      10  -5     0    0    0 S  0.0  0.0   0:00.00 xenbus             
   49 root      10  -5     0    0    0 S  0.0  0.0   0:00.05 kblockd/0          
   50 root      20  -5     0    0    0 S  0.0  0.0   0:00.00 kacpid             
  169 root      20  -5     0    0    0 S  0.0  0.0   0:00.00 cqueue/0           
  171 root      20  -5     0    0    0 S  0.0  0.0   0:00.00 ksuspend_usbd      
  174 root      10  -5     0    0    0 S  0.0  0.0   0:00.00 khubd              
  176 root      10  -5     0    0    0 S  0.0  0.0   0:00.00 kseriod            
  198 root      15   0     0    0    0 S  0.0  0.0   0:00.16 pdflush            
[climber@lode ~]$ vmstat
procs -----------memory---------- ---swap-- -----io---- --system-- -----cpu------
 r  b   swpd   free   buff  cache   si   so    bi    bo   in   cs us sy id wa st
 0  0  38984  44296 196536 356332    0    0     2     6    3   29  4  2 93  0  0

And from vmstat's man page FWIW:

Code:

   CPU
       These are percentages of total CPU time.
       us: Time spent running non-kernel code. (user time, including nice time)
       sy: Time spent running kernel code. (system time)
       id: Time spent idle. Prior to Linux 2.5.41, this includes IO-wait time.
       wa: Time spent waiting for IO. Prior to Linux 2.5.41, included in idle.
       st: Time stolen from a virtual machine. Prior to Linux 2.6.11, unknown.

Thinking · 03-18-2007, 05:31 PM

@sahithi
the values in /proc/stat don't decrease
see: http://www.linuxhowtos.org/System/procstat.htm

as the link describes the values in /proc/stat identify the amount of time (that's why it doesn't decrease, cause time continues

) the cpu has spent in what kind of "work"

for example, the first /proc/stat line shows like this at t0
cpu user nice system idle iowait irq softirq (dont know)
cpu 736324 49874 581102 298162832 3012015 120311 0 0
a few seconds later (t1):
cpu 736334 49874 581102 298162971 3012015 120311 0 0
differences:
cpu 10 0 0 139 0 0 0 0
all numbers together at t0 are: 302662458
a few seconds later (t1): 302662607
difference (t1-t0): 149

the difference 149 is 100% cpu time
so it seems the cpu has spent 93.28% (139 time unit's (seems to be USER_HZ)) in cpu idle
and 6.72% in user mode

i think the problem with top is that top causes cpu load too, which tapmers the cpu values

hope this helps

sahithi · 03-19-2007, 12:07 AM

Ha..

"Aggregation " is what I missed in my interpretation.

Thanks a lot..

syg00 · 03-19-2007, 12:59 AM

Quote:

Originally Posted by Thinking

i think the problem with top is that top causes cpu load too, which tapmers the cpu values

Mmmm - see this discussion on the observer effect.
At the default 3 second interval, I would say top is the lightest of the samplers - they all probe /proc after all.