Netgate Discussion Forum
    • Categories
    • Recent
    • Tags
    • Popular
    • Users
    • Search
    • Register
    • Login
    Introducing Netgate Nexus: Multi-Instance Management at Your Fingertips.

    Specifying Queue Length (maybe >100) causes 100% Ipv4 Gateway packet loss

    Scheduled Pinned Locked Moved Traffic Shaping
    1 Posts 1 Posters 91 Views 3 Watching
    Loading More Posts
    • Oldest to Newest
    • Newest to Oldest
    • Most Votes
    Reply
    • Reply as topic
    Log in to reply
    This topic has been deleted. Only users with topic management privileges can see it.
    • C Offline
      CasWanda
      last edited by

      Hi all!

      I have decent internet (Dual stack, symmetrical 2Gb/s XGSPON) and was playing around with FQ_CODEL in the Limiters and have followed Netgate's guide for the setup. I have done everything besides specifying the Queue Length, I left that blank. While tinkering with the performance I noticed that (as far as I understand) the Queue Length defaults to 50 while the recommendation is to have a much higher number, especially if approaching/beyond Gb/s speeds. First question is, do I need to worry about this when bufferbloat tests show very good latency under load (+/-1ms at 950Mbps)?

      Then, because I am tinkering with this, I specified the Queue Length to 3000 and then my whole internet stopped working. 100% packet loss on the Ipv4 Gateway, IPv6 did not show any issue, though I lost full connection to the interwebs. Looking around it seems something related to the dummynet or the whole Limiter not being initialized when specified with such Queue length. I read some, already "solved" questions about this in the forums talking about anything >100 on the QL would result in this issue but that it was solved. Maybe the issue is here again? I have just updated to the current stable pfsense version (26.07), the version before this one also had the issue (I don't remember exactly which version). I am also using a Netgate 6100 box.

      So, since bufferbloat tests show very good latency should I worry about this at all? LLM's say that a 3000 or more for the QL would absorb a bit more of bursty traffic before dropping packets unnecessarily. But while using this at 50 QL (defaults) I don't notice anything. But I am trying to learn about this and would like to tinker and actually have things working in an optimal way. Maybe for my specific connection, FIFO would already be optimal since I then get +1 to +4 ms when loaded, and this at 2.18Gb/s on the WAN.

      What do you guys think and am I missing something where the Queue Length is not really doing much when the Parameters of the FQ_CODEL Scheduler is already taking care of everything?

      Thank you for your attention, any input or discussion about this is greatly appreciated!

      1 Reply Last reply Reply Quote 0
      • First post
        Last post
      Copyright 2026 Rubicon Communications LLC (Netgate). All rights reserved.
      Privacy Policy · Cookie Policy