Ad Widget

Collapse

Trigger dependencies (multiple % CPU load)

Collapse
X
 
  • Time
  • Show
Clear All
new posts
  • EmmBee
    Junior Member
    • Jan 2019
    • 23

    #1

    Trigger dependencies (multiple % CPU load)

    Hello,
    i have 3 triggers on the one item.
    BTW It reads precent utility from windows server from WMI by bash script.

    Question is.
    How to make dependence between 3 triggers. for example when CPU load is more then 90% doesnt make any sense to activate trigger with 70% or 50%.
    I need see only last alert. How to make cascade? If 90% ...i dont need 70 and 50. If 70% i dont need see 50%..... (lot of alerts with no sense)


    Triggers for one server:
    {Server_windows_reading_from_WMIroc_percUtility.sh[{HOST.HOST},"_Total"].avg(5m)}>90
    {Server_windows_reading_from_WMIroc_percUtility.sh[{HOST.HOST},"_Total"].avg(5m)}>70
    {Server_windows_reading_from_WMIroc_percUtility.sh[{HOST.HOST},"_Total"].avg(5m)}>50

  • doum
    Member
    • Oct 2019
    • 35

    #2
    Originally posted by splitek
    You can add additional conditions to this triggers.
    trigger 1: >50 and <70
    trigger 2: >=70 and <90
    trigger 3: >=90
    This will work without trigger dependency, and imo. be more readable
    the problem of this solution is that when the cpu usage go from 60% to 80%, you receive a resolved for trigger1. it's not logical the problem is not resolved

    Comment

    • doum
      Member
      • Oct 2019
      • 35

      #3
      Originally posted by splitek
      You can block "resolved" for every trigger by setting option "Recovery expression" in trigger definition (ie. <50 - all events will stay until CPU drops below 50%).

      Here are example when dependency have sense:

      In my opinion in CPU usage monitoring, is no dependency at all.
      but if you do that, you will have two problems open. one for warning, one for critical

      Comment

      • doum
        Member
        • Oct 2019
        • 35

        #4
        Originally posted by splitek
        If you want one event then you need only one trigger. If you have more triggers then you have more events.
        Like I said in CPU usage monitoring, there is no dependency at all. Setting it has no sense. Lets look...

        Dependency (A -> B -> C):
        Trigger for A (>50) = no dependency
        Trigger for B (>70) = depends on A
        Trigger for C (>90) = depends on B
        If A is triggered, you will not get alarm for B (per dependency setting), and you will not get an alarm for C because of nested logic.

        Dependency (C -> B -> A):
        Trigger for A (>50) = depends on B
        Trigger for B (>70) = depends on C
        Trigger for C (>90) = no dependency
        At first look you may think "this is what I need", but... If A is triggered, you will get alarm, next CPU go up and raise B - you will get alarm for B. Again CPU go up and raise C and again you get an alarm for C.
        and the second is good ! the severity of the trigger is not the same
        if you go from A to C you have to know no?

        Comment

        Working...