check_ncpa get wrong alert from CPUs

[sacom01]

hi team,
i use Check_ncpa for check cpu. And our server has many CPUs.

when i set 'aggregate=avg' it got wrong information, the server come to more 80-100% but it show more than 60% only.
when i set 'aggregate=max' it got Critical always, seem it get info from highest cpu.
and if i delete that argument, it will get info from all cpus.

what can i do for get and alert exact info from CPUs.

thanks.

[ssax]

AVG will get you the average of all CPUs combined.

MAX will use the highest value of all CPUs.

MIN will use the lowest value of all CPUs.

Please run that check command again without arguments (so I can see all CPU percentages) and then attach the text version of it so that I can use the values to see if it's calculating it properly.

Thank you!

[sacom01]

pls find attach file for details.

[ssax]

You'll want to use the aggregate=avg:

/usr/local/nagios/libexec/check_ncpa.py ... egate=avg'

Please note that the timing of when the top checks vs when NCPA checks will be different, you are likely rarely going to see them match.

What NCPA version are you running on the remote system? What OS/version is it?

[sacom01]

NCPA Agent_version was ['2.2.1']
os level: AIX 7.2.3

[ssax]

What is the full output of this command?

Code: Select all

/usr/local/nagios/libexec/check_ncpa.py -H X.X.X.X -t 'yourtoken' -P 5693 -M cpu/percent -w '20' -c '40' -q 'aggregate=avg'

[sacom01]

i told you ssax, avg get wrong info.
our system ab 80-100% performance but it show about more than 60% only

[vtrac]

Hi,
Hope you are having a good day!! ...

Let try this, we will try to get the average CPU on your NCPA remote.
I am assuming that your NCPA agent is windows.

Please open a Nagios XI command prompt, then run the steps below.

1) Let get number of CPU cores in your remote NCPA agent.
NOTE: x.x.x.x is your remote NCPA agent, and "yourTOken" is your NCPA's token defined in "ncpa.cfg" file

Code: Select all

curl -k "https://x.x.x.x:5693/api/cpu/count?token=yourToken"

Example outputs:

Code: Select all

{
    "count": [
        [
            8
        ],
        "cores"
    ]
}

2) Now, sum (total) of CPU percent of each CPU:

Code: Select all

curl -k "https://x.x.x.x:5693/api/cpu/percent?token=yourToken" | egrep "[0-9]" | awk -F',' '{sum+=$1;}END{print sum;}'

Example outputs of step (2):

Code: Select all

106.3

3) Now, take the output of step (2) and divide that with number of CPU cores in step (1).
This will give you the average CPU percent.

Example: average CPU/percent

Code: Select all

106.3 / 8 = 12.2875 

Does this match your other average?


Best regards,
Vinh

[sacom01]

well, let me tell you details, it's too complex

Our system have both Redhat and AIX servers
when i run the command and check with Nagios XI, it match number between avg in command and Nagios XI.

When i check avg in command with TOP and TOPAS command on servers :
REDHAT : avg in command > number in TOP command (eg : 20% > 3%)
AIX : avg in command < number in TOPAS command (eg : 20% < 45%)

Now i'm very confuse and do not believe with which number (

by the way, Vinh pls check inbox.

[vtrac]

Hi Sacom01,
CPU changes so fast that it is very hard to get the exact number.
In your case, looks like your system is very large, with so many CPU's and some not even being used (many are 0%).

I noticed in your opening page, your system is set to monitor every 1 minute.
I would suggest that you change that to "5 minutes" instead since your system is large. Monitor every minute does not make sense since many are not even used ( at 0% ... based on picture).

The issue we want to watch for is that CPU stay at 90+% at a long period.

Please let me know if there is anything else that we can do to help!!


Best Regards,
Vinh