Ad Widget

Collapse

Proxy - IPSLA with around 30.000 items

Collapse
X
 
  • Time
  • Show
Clear All
new posts
  • M2r12n0
    Junior Member
    • Aug 2021
    • 14

    #1

    Proxy - IPSLA with around 30.000 items

    Hello guys,

    Any suggestion about this? I would need to deploy a discover rule to a router with around 300 IPSLAs sentences, and those generate something like 30.000 items.

    First try, the proxy crashed (8GB RAM, 4 vCPU).

    Thank you, I appreciate it.
    M
  • um3n
    Member
    • Aug 2021
    • 47

    #2
    Originally posted by M2r12n0
    Hello guys,

    Any suggestion about this? I would need to deploy a discover rule to a router with around 300 IPSLAs sentences, and those generate something like 30.000 items.

    First try, the proxy crashed (8GB RAM, 4 vCPU).

    Thank you, I appreciate it.
    M
    Did you check the requirements from the documentation?

    Maybe it would be helpful, if you could provide some systemlogs and performance counter.

    Comment

    • M2r12n0
      Junior Member
      • Aug 2021
      • 14

      #3
      Hello !!
      Thank you for the feedbacks both !!

      The truth is I found in the requirement page information about database size, but that is not related (from my understanding) to the performance of having a lot of items to monitor.
      So, for example, You could have 20 years of historical information with 500.000 items to monitor every 10 hours, and another scenario would be having only 3 months, but with 10.000 items every 30 seconds.
      Probably (I did not do the math) the space in the DB will be similar, but the CPU and memory consumption is clearly very different.

      So, I just wondering how to calculate this consumption, I mean, who enforces the rate limit (throttle) of every action so I can predict if the 10.000 requests are going to be redistributed in time How !? Because that is what It will give Us the performance.

      Thank you
      Mariano

      Comment

      • um3n
        Member
        • Aug 2021
        • 47

        #4
        If you want to see it from the os-perspective, there are some futher things to think about... ulimit, connections and filehandles and so on

        The requirementspage has also a table included, where you can see some recommendations for the hardwaresizing. You could also think about a proxy infrustructure, so your zabbix-server is not so loaded with actions and polling. In most cases you will use a proxy for communicating in different subnets, but it also helps with your performance.
        Also Zabbix 6.0 now has high availability, maybe this could help as well.

        I'm sorry, that there isn't an easy answer to your question - at least from my perspective !

        Comment

        • cyber
          Senior Member
          Zabbix Certified SpecialistZabbix Certified Professional
          • Dec 2006
          • 4807

          #5
          If your proxy crashed, maybe there is some pointers in log, what caused it to crash? Ran short of some resources/processes etc? If you increase those, maybe it manages... Need more memory or preprocessing threads or .. .things..
          30k items at once is big amount to chew... You could try to reduce the amount of created items... 100 parameters for one thing seems quite excessive (30k /300)...

          Comment

          • M2r12n0
            Junior Member
            • Aug 2021
            • 14

            #6
            Thank you, guys! Working on it.
            I hope to have access today to the proxy, it is not under my local administration.
            I will give you feedback after checking the logs.

            Comment

            • M2r12n0
              Junior Member
              • Aug 2021
              • 14

              #7
              Well, cache problems. I just increased it.
              I took off anyway some items, I will try them in some days.

              Comment

              Working...