DX Infrastructure Manager

  • 1.  Probe netconect

    Posted 10-20-2021 01:59 PM
    How many tests does the netconect probe support?


  • 2.  RE: Probe netconect

    Broadcom Employee
    Posted 10-20-2021 02:12 PM
    Explain what you mean by 'test'.


  • 3.  RE: Probe netconect

    Posted 10-20-2021 02:21 PM

    Hello,

    I'm sure there's a limit document somewhere but it will also depend on the timeout value you configured, the failed interval and the interval. If it's a very long timeout and a lot of retry during that period you could face the issue that you will not be able to finish querying all your node and it will already be the time to do the next interval.


    This is the alarm that you will see:

    You will not see the same behavior when everything is reporting OK and when a lot of things will stop responding,

    What I mean is you could be able to ping 2000 devices without any issues when they all respond. But as soon as there's an outage and you can't reach 100 or 200 devices out of the 2000 depending of the way you configured the timeout value and the interval you could not be able to complete querying all the devices.

    You can also play with the ping / service thread value but there's a limit of 100

    At least, that's how I interpret the way this probe is working. I could be wrong

    Thanks




  • 4.  RE: Probe netconect

    Broadcom Employee
    Posted 10-20-2021 04:26 PM
    Hi Jose,

    In the past, the development team tested net_connect and stated that it supports around 800 profiles with 5 min interval (with the default values for retry, timeout, etc.).

    If you encounter net_connect monitoring issues such as connection to <ip> failed, profiles that don't finish, etc., you can split up the profiles across additional hubs/robots using net_connect if you start to have performance issues with the probe like false alarms or no QOS data, and/or alarms such as the following:

    - Connection to '10.x.x.xx' (ping) failed (profile: 10.x.x.xx)
    - An increase in the number of profile timeouts
    - Profile failed to execute in scheduled time interval
    - etc.

    In the release notes, it states for a very old version that performance improvements were done to handle 5000+ profiles. That said, you can make a post to the community but I've never seen that high of a number, at least not on a single machine.

    We do have some customers that have thousands of net_connect profiles being monitored from a single hub/robot but it always depends on a number of factors such as available system resources, and probe configuration such as monitoring interval, number of threads, timeout settings, etc.

    Your best bet is to set this up in your test environment to do some extensive performance testing (drag and drop to configure a large number of profiles), load up the probe and try to overload it, to see if you can come up with a baseline/benchmark.

    Steve

    ------------------------------
    Support Engineer
    Broadcom
    US
    ------------------------------