2 Replies Latest reply on Sep 18, 2019 10:37 AM by Wlacroix

    Configuring sflow on HP Switches

    andrew7

      Good Afternoon,

       

      In the next few weeks are deploying NTA into the business, and we have a range of HP network switches that require configuring for sflow the different models of HP switch that we need to configure are 2530, 2510, 4208vl, 3500yl, 2810, 2650.

       

      From watching this Solarwinds labs video https://www.youtube.com/watch?v=OhaOKKJRCLk it suggests only the following command is required to setup sflow for NTA;

       

      sflow 1 destination x.x.x.x 2055 (x.x.x.x being the Solarwinds server).

       

      Could anyone please confirm if this is all that is required for sflow to work with Solarwinds on HP devices?

       

      Regards,

       

      Andrew

        • Re: Configuring sflow on HP Switches
          Wlacroix

          Andrew,


          We have this running in our environment across 80+ HP\Aruba switches, I would be happy to guide you through the setup process.

           

          Here is all my sflow script stuff. We primarily run Aruba 2930F, HP 5412zl, HP5406zl, HP3524YL, HP3548YL and several other models.

           

          sflow 1 destination IPADDRESS:PORT

          sflow 1 polling PORT \ time

          sflow 1 sampling port \ samples

           

          EXAMPLES:

          sflow 1 destination 10.100.100.198 2055

          sflow 1 polling 24 60

          sflow 1 sampling 24 4096

           

          sflow 1 destination 10.100.100.198 2055

          sflow 1 polling all 60

          sflow 1 sampling 2-6, 15 4096

           

          In most cases set your polling time to 30 or 60 seconds, longer is usually OK unless you need super granular stuff

          In most cases set your samples to 8192 this is the default, I have mine set to 4096.

          If you want more samples and quicker set them smaller, please be aware that this puts stress on a switch and may increase the drop rates.
          Also note that the more switches you have with smaller samples will put stress on your egress queues with small packets of almost useless data

          As an example, 80+ switches cause roughly 1/4 a megabit of traffic to my orion server. And we dont monitor every port. Just key links from Head office to Branches or other switches in our environment.


          If you run:
          sho sflow 1 sampling-polling from a config you will get an out put. Here is one of mine:

          SWITCH NAME (config)# sho sflow 1 sampling-polling

           

          SFlow Sampling Information

           

                  | Sampling                    Dropped | Polling

            Port  | Enabled      Rate Header    Samples | Enabled Interval

            ----- + -------  -------- ------ ---------- + ------- --------

            B1      Yes(1)       4096    128          0   Yes(1)        60

            B12     Yes(1)       4096    128          0   Yes(1)        60

            B14     Yes(1)       4096    128          0   Yes(1)        60

            B15     Yes(1)       4096    128          0   Yes(1)        60

            B19     Yes(1)       4096    128          0   Yes(1)        60

            C17     Yes(1)       4096    128          0   Yes(1)        60

            G22     Yes(1)       4096    128          0   Yes(1)        60

            K1      Yes(1)       4096    128       2522   Yes(1)        60

            Trk1    Yes(1)       4096    128          0   Yes(1)        60

            Trk2    Yes(1)       4096    128          0   Yes(1)        60

            Trk5    Yes(1)       4096    128       2627   Yes(1)        60

            Trk6    Yes(1)       4096    128          0   Yes(1)        60

           

          You can see I have some sample drops. I am working on why they are there today. I do have an egress issue on trunk2 that will be repaired when we move it to 10g (long story) so im not panicking on this.

          Best Regards,
          Wally