Wrong traffic graph?

  • Hi!

    I tought to report this one…
    See image wrong_graph_1.JPG.
    I did simple file copy from one net to another. Why such difference?
    On OK_graph_1.JPG I did just the opposite. From other LAN to my LAN file copy and it is showing OK.
    Is this because LAN is configured as LAGG and wrongly interpretated by graph or what?


  • I know of one known issue with the Graphs: physical and "logical" Interfaces are not actually seperated the way they should be. For example, if your WAN runs on VLAN7 on the physical Interface em0, and you use OPT1 without VLAN tagging on em0 (to access the modem's admin Interface), then the OPT1 traffic graph will Show all traffic on the physical em0 Interface - not just the untagges one, as you would expect, but also the WAN (VLAN7) traffic.

    You can also see another issue with the traffic graphs. They indicate that the traffic drops to zero for a short period every 30 seconds or so. This is definitely not true; the traffic runs smoothly without any breaks.

    Both issues have been present for quite some time. I noticed this since 2.0, but it may have been been present much earlier - I must admit that I never spent much time staring at the traffic graphs. They arr nice on m0n0wall, because there they are all you got, but on pfSense, I prefer to look at the RRD graphs (they accurately seperate the traffic by logical interface, as expected) or vnstat2 (haven't checked if vnstat2 works correctly in that regards, though…). With the softflowd package, pfSense can even generate "industrial strenght" Netflow datagrams (which are, due to the nature of Netflow, not strictly real-time).

    One might also regard the aggregation of traffic on physical interfaces a "feature" instead of a bug. Someone might be more interested to see if a physical interface gets saturated. However, even in such a case, it might be nicer to be able to choose between "total em0 traffic" and "untagged OPT1 traffic only".

    Whatever. Given the long-time presense of these two bugs, and the lack of "public uproar" so far, I suspect that about everyone cares very little about that...after all, the traffic graphs are very nice eye-candy on the dashboard, but for erious digging, other tools are more apt anyway.

    Of course, noone would mind if some would volunteer to fix these issue. ;-)

  • I can confirm this on 2.1.1-PRERELEASE (amd64) - built on Wed Mar 19 15:06:47 EDT 2014  - FreeBSD 8.3-RELEASE-p14.

    See attachment.

  • I did as suggested here:

      var ifout = parseInt(t[2]/2, 10);   // number of bytes sent by the interface

    Seems to have solved it.

  • The drops to 0 when the counter wraps. On 64 bit systems, that's very rarely. On 32 bit, it's pretty frequent if you're pushing a decent amount of traffic.

    Double traffic on VLAN SVG graph is: https://redmine.pfsense.org/issues/3314

Log in to reply