Traffic graph reports double the real amount on some interfaces.



  • Hi,

    I have a pfSense box here with three physical interfaces, em0 em1 and bge0. em0 and em1 both have two VLANs each, into the same Cisco switch. The VLANs on em1 (WAN and WAN2) report the correct amount of traffic in the dashboard, but the VLANs on em0 (both local networks) show double the real bandwidth usage. Is this something anyone else has seen before?

    Thanks.



  • OK, an update on this one. All the VLAN interfaces show double the amount of "out" traffic and are accurate on the "in" traffic.



  • And I thought it was my messed up config…

    I can confirm this on vge interfaces with VLANs on top.
    Using 2.1 on nanobsd (4g).



  • same double the bandwidth,
    this really sucks when i use that to monitor user traffic

    running installed iso on physical machine, used pfsense updater…



  • Nice to hear I'm at least not alone on this one. I'll report back if I manage to find a fix. Thanks guys.



  • Just as a follow up - I see the same behaviour on an ALIX board when using VLANs.
    It seems to be independent of the interface type (VR on ALIX) and occurs with virtual interfaces only.

    I'll try to point the developers to it.



  • I see the same issue here. I also noted that shaper queue status is wrong/too high, sometimes WAY off (i.e. a vlan with 3mbps MAX available from WAN showing 24mbps out to LAN).

    I noted this RRD issue in the previous 2.x version as well, but it has gotten much worse and spread to queue status in 2.1



  • I stumbled onto this while setting up a limiter.  Attaching a screenshot of the traffic graphs.  Notice that the Guest interface is showing double the Fairpoint interface; VLANs 6 and 4 respectively on em0.  I have the debug menu turned on on this Roku device and it also senses a 1.5MB network (the limiter value, accurate on the Fairpoint interface).

    The directions would be reversed between the two, but I did a line-saturation upload on the Fairpoint interface and the 'Out' value is correct there.  That's the interface with the current gateway (top tier in the gateway group).




  • Same issue VLANs interfaces traffic is double, i was also thinking about configuration i might configure wrong. any one has nay idea why pfsense behave like that is its a bug.



  • I am having an identical issue. Has any one found any information regarding this?

    Thanks,



  • Hi,

    I have same problem … here



  • Affected here too.

    • v2.1-RELEASE (i386).

    • Running in an ESXi 5.1 VM, using VMXNET 3 vNICs (vmx3f#).

    • There are no VLANs configured in pfSense directly, but some vNICs use vSwitches that have a VLAN tag.



  • @ab0tj:

    OK, an update on this one. All the VLAN interfaces show double the amount of "out" traffic and are accurate on the "in" traffic.

    Mine's the opposite - 'in' is being doubled, while 'out' is fine.

    For now, I've changed the line in /usr/local/www/graph.php to:

    var ifin = parseInt(t[1]/2, 10);    // number of bytes received by the interface
    

    … in the below code block:

    var t = obj.content.split("|");
    var ugmt = parseFloat(t[0]);  // ugmt is an unixtimestamp style
    var ifin = parseInt(t[1], 10);    // number of bytes received by the interface
    var ifout = parseInt(t[2], 10);   // number of bytes sent by the interface
    var scale;
    

    There might be a better/correct bandaid, but that'll do for the minute.



  • Just to say I'm also seeing this.

    2.1.2 amd64 on an Intel 2500CC platform with no Vlans.

    If I look at my traffic graphs for WAN and LAN then the 'in' of both is correct, but the 'out' of both is exactly double what it should be.



  • Same here. No vlans, 2.1.2 amd64 em0 card.

    WAN card is correct, LAN card is double on out values.



  • Hey

    Same here. Using BLKD2500CCE, no VLANs.

    Any other ideas instead of edit graph.php?

    Bandwidthd and Vnstat2 also shows double amount of traffic on lan interface.



  • I also have this issue. This has been discussed several times, see for example this https://forum.pfsense.org/index.php?topic=75343.0..

    Probably this is a known error (https://redmine.pfsense.org/issues/3314). It seems that the error lies in the OS and developers discuss that this might be fixed with the upcoming version 2.2.

    -flo-



  • Same issue here using the built-in ethernet ports on our super micro server - LAN shows twice out traffic as WAN in traffic. No VLANs.



  • Also seeing this problem - testing upload bandwidth with iperf, the wan graph is showing double the outgoing traffic that the lan graph is showing coming in.



  • I'm seeing this exact issue in the latest 2.2 snapshots (screenshot attached). It also seems to be directional: when I send traffic from a server in my DMZ to a machine on my LAN, the DMZ registers the correct level but the LAN doubles he traffic. When I send traffic in the reverse direction, the opposite occurs (again, see screenshot below).

    The server in my DMZ is running off a 2-port LACP connection, and both networks are using VLANs (which seems to be the factor that causes the issue).

    Looks like the redmine bug has already been updated to reflect the fact that this isn't fixed in 2.2, but FYI.

    https://redmine.pfsense.org/issues/3314



  • Banned

    Not having this issue in 2.1.4 just for info.






  • @asayler

    Didn't you uncheck the box "Get a free copy of your private data in the iCloud"  ;D :P



  • I am seeing the same thing on my system. I am not using VLAN's at all. I am on Version 2.1.5 and my hardware is NetGate. If I look at the RRD graphs everything is accurate. It is only on the Traffic Graphs.

    Thanks



  • Also seeing this on 2.1.5 on an Atom D510 based system with onboard LAN (em1) and WAN (em0) ports for an traffic on the OUT of a port. Specifically, when running a speedtest via speedtest.net:

    On the download test, the IN on WAN shows a match to the speedtest while the OUT on LAN shows double.

    During the upload test, the IN on LAN shows matching speed while OUT on WAN shows double.

    Screen grabs of speediest and bandwidth graphs attached.




  • @asayler:

    The server in my DMZ is running off a 2-port LACP connection, and both networks are using VLANs (which seems to be the factor that causes the issue).

    No it's not. At least not the only cause. There are several reports in this thread alone that state they are not using VLANs…

    Maybe we can find a pattern if we start to give more detailed info of the affected (and non-affected) interface combinations?
    pfSense: 2.1.5-RELEASE (amd64) (and 2.1.4-RELEASE (amd64) and 2.1.3-RELEASE (amd64))
    WAN: Intel 82579L (on Intel DQ77KB)
    LAN: Intel 82579LM (on Intel DQ77KB)
    VLAN: No
    Symptom: outgoing traffic reported 2x in both directions (upload and download)


  • LAYER 8 Netgate

    It's a known bug in the underlying FreeBSD.



  • @Derelict:

    It's a known bug in the underlying FreeBSD.

    Can you give a link to the bug?
    It would be good to fix up little crud like this. It puts off people new to pfSense when something like the traffic graph shows an obviously wrong figure.


  • LAYER 8 Netgate



  • I have the same issue.  WAN shows correct speed but my VLAN does not, unless it is running at a lower speed of say 30 Mbps.

    Anyone also notice labels don't match!




  • Anyone also notice labels don't match!

    Not sure which labels you are referring to.
    VLAN_4 interface has a lot of traffic going out, and host (client) IP 10.0.8.20 is receiving a lot of traffic in. That is what happens when something is downloading - pfSense sends a lot of traffic out the LAN/VLAN… and the client receives it in.
    (Of course there is the double thing happening there in the graph plot also)



  • But this is exactly the point (On internal VLAN)

    The graph you saw in previous post was for inbound (downloaded) traffic that you are seeing (albeit doubled).

    The table on the right is correct, but the graph and top left labels are reversed to the table

    To help illustrate this I waited until I had a distinctive spike that also did not obscure the chart.  WAN is spot on, so suspect someone reversed the table for LAN internal connections, but forgot the labels in top left?

    Do you see the label mismatch?




  • That what I was trying to explain. "In" and "Out" are done from the point of view of the device concerned.
    Out of pfSense LAN goes in to a LAN client.
    Out of a LAN client comes in to pfSense LAN.
    So you expect to see "opposite" because one side is the sender and the other is the receiver.



  • OK, Understand.

    I had assumed they were directly the same.  Now that I know, it makes sense but could have been clearer on the chart what was meant.

    Is there anyway to make the chart page bigger and therefore longer - I am interested in the table part on the right side.

    Thanks.



  • At the moment that table is hardcoded to 10 rows - both in /usr/local/www/status_graph.php and bandwidth_by_ip.php
    It's only software - there could be a parameter box to select how many rows to display in the table.



  • Will this be fixed or?

    I am running Supermicro server with no vlans. 2.1.5 - very annoying :)


Log in to reply