[SOLVED]Periodically (random) low throughput, no apparent reason - Help!



  • Hey guys!

    So I had my PFSense router running for a while now, but now I am experiencing stuff like this:

    https://www.dropbox.com/s/oyrchaqmrlip26v/Screenshot 2016-03-08 14.07.09.png?dl=0

    It happens about every 5 minutes and doesn't last long, but it's so persistent that it's seriously hindering the overall throughput.

    What I did so far was looking at different system stats according to this:
    https://doc.pfsense.org/index.php/High_Load_Troubleshooting
    and this:
    https://doc.pfsense.org/index.php/Low_Throughput_Troubleshooting

    Here is what I tried:

    • During the times of extremely slow net my Interrupts go to about 2-5% so that is probably not it.

    • The general Hardware load is always low, between 0-10% (5-20% for memory) no matter what I do.

    • Traffic Shaping is off

    • The box has a Realtek RTL8111/RTL8168 NIC (2x) which is on direct throughput in ESXI

    • No Duplex mismatch, tried the Hardware tuning - didn't make a difference

    • Checked for slowness between PFSense and Modem, but the response times go up noticeably even when I ping PFsense directly (Tried with different WAN connections as well)

    • Turned it off and on again

    If you need any more info please don't hesitate to ask.

    Thanks for reading, hope you can point me in the right direction.



  • There are a few things to look at:

    • Monitor the traffic graph when the pings are high to verify it's not due to bandwidth usage.

    • It looks like your PFsense is virtualized (ESXi 6?), is the box powerful enough to handle your firewall along with everything else?

    • Have you tried the VMXNET3 driver?  VMware tools installed?

    • Replace all your cables with prefabbed Cat6

    • Most people will tell you to replace your NIC's and go with Intel (PCIe)

    • If you have a spare box, re-install on bare-metal and see if the issue remains

    • Once you've ruled out everything on your end, use an app like visual traceroute or ping plotter to document exactly which hop is producing the latency and take the evidence to your ISP.



  • 1 - Will do.
    2 - Yes, definitely. I only have a (shut down most of the time) windows 7 host besides PFSense.
    3 - Yes and yes
    4 - Everything in my cabinet is patched with Cat6 (Cat 7 through the walls between sockets and Patchfields)
    5 - Replacing the NICs is not an option, would have to change the entire device (Using a compact ZBox as a host here because we had it available)
    6 - I'll try this, thanks.
    7 - Same.

    Thanks, I'll post again when I get to do all this and report my findings.



  • What about a closer target to ping instead of something that is probably 10+ hops away with 10+ variables.



  • Turns out something was literally pulling 27mb/s and slowed everyone down. The problem is fixed now.

    Also @Harvy66
    I pinged from my Laptop to the Router (1 hop) and I pinged from my Laptop to 8.8.8.8 (9ish hops), also from PFSense to 8.8.8.8 - Usually the values are averaged on 1ms, 13ms and 11ms respectively, during the extreme load the averages were like 6ms, 400ms and more. Thanks for your suggestion though, now just to find out what's around that's closer but still connected via WAN  ;)



  • That bandwidth graph is useful when trying to correlate throughput and latency. If you configure PFSense correctly, you can dramatically reduce ping spikes.


Log in to reply