Traffic Graph Negative Numbers



  • Hi,

    In the attached screenshots, are the negative numbers on the traffic graphs due to the spiking in the graph? These results were gathered with an iperf test between two linux servers (one on LAN and one on LAB_NETWORK).

    The commands I ran were iperf -c 10.0.0.248 -f M -d -t 360 on the client and iperf -s -f M on the server.

    Thanks,



  • That looks a lot like a 32-bit counter overflow, resulting in negative numbers when the true number is over 2Gbps, given that you are running tests that achieve that sort of bandwidth. I seem to remember this being discussed a while ago, when multi-Gb interface speeds became more common. But I can't remember if it got fixed/enhanced recently in GitHUb.



  • That would make sense from what I was seeing. Is it safe to assume this will not affect anything other than the graphs? This is a home pfsense install that will not be running many packages but will route traffic between the interfaces at gigabit when needed. The WAN bandwidth is only 50/30.


  • Netgate Administrator

    But shouldn't the real figure but ~700Mbps?

    What NICs are you using to achieve >1Gps?

    Steve



  • They are gigabit nics - E1000 - ESXi 5.5

    I uploaded a video of the graphs on LAN as well as a iftop of em1 on the router. Iftop shows that the speed does not drop off like in the graph so why do the graphs show it spiking so badly?

    Youtube Video

    Thanks for your help.


  • Netgate Administrator

    So those numbers are all wrong? It cannot possibly be over 1Gps, unless you have a LAGG set setup.

    Steve



  • even with LAGG groups or LACP you can only get maximum of 1 NIC speed no?



  • There is nothing special configured with the interfaces. They are single link gigabit interfaces. The only place it shows above 1 gig is on the graphs as well as showing the spikes/drop offs.


Log in to reply