Ad Widget

Collapse

Interpreting system[procload] values

Collapse
X
 
  • Time
  • Show
Clear All
new posts
  • dm0nkz
    Junior Member
    • Jan 2005
    • 9

    #1

    Interpreting system[procload] values

    In the Zabbix reference manual, the key "system[procload]" is

    a) described as the load average for 1 minute.
    b) values on the graphs are NOT in percentage.

    When i look into the graphs i see that the values on the vertical-right side of
    the graphs are 0.70, 1.40, 2.10, 2.80, 3.50, 4.20 and so on.

    How do i interpret this values? what do they mean?

    Thank you.

    dm0nkz
  • bennethos
    Member
    • Jan 2005
    • 31

    #2
    if you do a 'top' you will get following :
    CPU states: 0.2% user, 0.2% system, 0.0% nice, 99.6% idle

    so the system is a cpu state between 0-100% ...

    grtZ
    Last edited by bennethos; 04-02-2005, 01:40.

    Comment

    • dm0nkz
      Junior Member
      • Jan 2005
      • 9

      #3
      Thanks for the reply... Is it the same in windows.
      I just cant figure out what the 4.3, 3.2, 5.0 figures represent.
      Like if you have a 2.4Ghz of CPU power what
      does 4.3 or 2.7 metric represent?

      Again thank you! Appreciate your input.

      dm0nkz

      Comment

      • PerroVerd
        Junior Member
        • Dec 2004
        • 11

        #4
        No, you are wrong

        Originally posted by bennethos
        if you do a 'top' you will get following :
        CPU states: 0.2% user, 0.2% system, 0.0% nice, 99.6% idle

        so the system is a cpu state between 0-100% ...

        grtZ
        No, you are wrong

        procload is about "load average".
        The numbers represent the count of processes waiting in the queue to be executed.
        in 1 min, 5 min and 15 min.

        If you do a top you will see
        load average: 1.55, 1.70, 1.51

        A value greater than 1 implies that one processor is not enough to take all the system load, a value greater than 5 for long time or so implies that your system has too much hugry cpu processes.

        If you want to check cpu usage you need to try the cpu[idle, nice, system, user] keys

        Best regards

        Comment

        • mcheedle
          Junior Member
          • Oct 2004
          • 3

          #5
          I needed a way to get the percentage of processor used, and hadn't realized that system[procload] didn't do that. I came up with a little script in PHP that parsed the information from TOP.

          *************
          #!/usr/bin/php -q
          <?php
          //1 iteration of TOP in batch mode
          $cpuraw = shell_exec('/usr/bin/top -b -n 1');
          $cpuraw = split("\n", $cpuraw);

          $cpupct = 0.0;
          for( $i=0; $i<count($cpuraw); $i++ ){
          //get rid of multiple spaces and lead/end spaces
          $current = preg_replace('/\s+/', ' ', trim($cpuraw[$i]));

          //use spaces as delimeters and break into an array
          $break = split(' ', $current);

          if( is_numeric( $break[8] ) ){
          $cpupct = $cpupct + $break[8];
          }
          }
          echo sprintf("%01.2f\n",$cpupct);

          ?>
          *************
          Copy the above data to a file (ie. cpuusage.php), add it to the system path, then make it executable.

          Create an entry in zabbix_agentd.conf for this:
          UserParameter=CPUT,cpuusage.php

          Comment

          • Rudolf
            Member
            • Dec 2004
            • 90

            #6
            windows processor

            For Windows machines I use the key cpu_util instead of system[procload]

            Comment

            • Lane
              Junior Member
              • Apr 2009
              • 7

              #7
              Processor Queue Length / CPU Count

              perf_counter[\System\Processor Queue Length]

              (reports threads waiting in queue for a processor)

              system.cpu.num[online]

              (reports number of processors)

              Processor Queue Length / CPU Count > 2

              could indicate a problem on most windows systems.
              Depends on the quality of the software and how busy the server is.

              You will still need to evaluate if the server is serving time critical data.

              Comment

              • nelsonab
                Senior Member
                Zabbix Certified SpecialistZabbix Certified Professional
                • Sep 2006
                • 1233

                #8
                Originally posted by PerroVerd
                A value greater than 1 implies that one processor is not enough to take all the system load, a value greater than 5 for long time or so implies that your system has too much hugry cpu processes.
                It could also mean you have a lot of process waiting on IO. If a process is blocked by IO then it will be listed as a process waiting for execution. Increasing IO performance can improve processor load in some cases. If you system supports it take a look at IO wait state statistics. Newer Linux distros and top support this.
                RHCE, author of zbxapi
                Ansible, the missing piece (Zabconf 2017): https://www.youtube.com/watch?v=R5T9NidjjDE
                Zabbix and SNMP on Linux (Zabconf 2015): https://www.youtube.com/watch?v=98PEHpLFVHM

                Comment

                Working...