Re: BGP Stuck Routes Operational Experience Survey

[Hank Nussbacher via NANOG <nanog () lists nanog org>]

On 07/04/2026 20:06, Bryton Herdes via NANOG wrote:

Hi,

When a zombie route appears in some stub or esoteric ASN, the damage is 
minimal.
But when a zombie route appears in someone as large as AS3356, that 
causes chaos.
Here is a traceroute for 128.139.7.0/24 - 30 minutes after a withdrawal 
was issued:

  1 vl-51-gw.uoregon.edu (128.223.51.1) [AS 3582] 116 msec 48 msec 48 msec
  2 10.252.19.140 48 msec 48 msec 48 msec
  3 10.252.10.177 48 msec 48 msec
    10.252.9.177 48 msec
  4 10.252.10.178 48 msec 48 msec 48 msec
  5 128.223.182.77 [AS 3582] 48 msec 48 msec 48 msec
  6 10.252.10.250 116 msec
    10.252.10.246 48 msec 48 msec
  7 eugn-oh-pe-02.net.linkoregon.org (207.98.127.249) [AS 3701] 48 msec 
48 msec 48 msec
  8 eugn-oh-noc-gw.net.linkoregon.org (207.98.126.13) [AS 3701] [MPLS: 
Labels 900249/100001 Exp 0] 48 msec
    1-1-c23-1.ear2.Sacramento1.Level3.net (4.53.203.181) [AS 3356] 48 msec
    eugn-oh-noc-gw.net.linkoregon.org (207.98.126.13) [AS 3701] [MPLS: 
Labels 900249/100001 Exp 0] 48 msec
  9 eugn-oh-pe-01.net.linkoregon.org (207.98.126.76) [AS 3701] [MPLS: 
Labels 900249/100001 Exp 0] 49 msec *  48 msec
 10 209.244.187.2 [AS 3356] 146 msec
    eugn-oh-pe-02.net.linkoregon.org (207.98.127.249) [AS 3701] 48 msec 
48 msec
 11 ae86.edge2.NewYork6.Level3.net (209.244.187.1) [AS 3356] 116 msec
    1-1-c23-1.ear2.Sacramento1.Level3.net (4.53.203.181) [AS 3356] 48 
msec *
 12 209.244.187.2 [AS 3356] 116 msec 146 msec *
 13 209.244.187.2 [AS 3356] 96 msec 96 msec 97 msec
 14 209.244.187.2 [AS 3356] 96 msec *  98 msec
 15 209.244.187.2 [AS 3356] 104 msec 106 msec *
 16  *  98 msec *
 17 209.244.187.2 [AS 3356] 96 msec 99 msec 99 msec
 18 ae86.edge2.NewYork6.Level3.net (209.244.187.1) [AS 3356] 96 msec *  
97 msec
 19 ae86.edge2.NewYork6.Level3.net (209.244.187.1) [AS 3356] 97 msec
    209.244.187.2 [AS 3356] 97 msec 165 msec
 20  *  213 msec
    ae86.edge2.NewYork6.Level3.net (209.244.187.1) [AS 3356] 98 msec
 21  *
    209.244.187.2 [AS 3356] 116 msec 100 msec
 22  *  *  97 msec
 23  *  97 msec 96 msec
 24  *  74 msec 74 msec
 25 209.244.187.2 [AS 3356] 74 msec
    ae86.edge2.NewYork6.Level3.net (209.244.187.1) [AS 3356] 81 msec
    209.244.187.2 [AS 3356] 75 msec
 26  *  *  75 msec
 27 ae86.edge2.NewYork6.Level3.net (209.244.187.1) [AS 3356] 134 msec
    209.244.187.2 [AS 3356] 74 msec
    ae86.edge2.NewYork6.Level3.net (209.244.187.1) [AS 3356] 82 msec
 28  *
    209.244.187.2 [AS 3356] 97 msec *
 29  *  98 msec *
 30  *  *  *

AS3356 appears here: 
https://www.thousandeyes.com/bgp-stuck-route-observatory/?asn=3356
How can I, a non-customer of AS3356, get them to fix their network so as 
to stop spawning zombie routes?

Thanks,
Hank


> Thanks for sharing the survey.
>
> In addition to responding, I thought I'd share a blog we wrote late last
> year detailing our operational experience with stuck routing:
> https://blog.cloudflare.com/going-bgp-zombie-hunting/
>
> --
> Bryton Herdes
> Principal Network Engineer
> AS13335 - Cloudflare
>
>
> On Mon, Apr 6, 2026 at 5:02 PM Iliana Xygkou via NANOG <
> nanog () lists nanog org> wrote:
>
> > Dear NANOG community members,
> >
> > We are researchers from Georgia Tech investigating the phenomenon of BGP
> > stuck routes (also known as "zombie routes")—prefixes that persist in
> > routing tables despite being withdrawn by their origin AS.
> > While recent research using BGP beacons (including our work presented at
> > IMC 2025 [1]) confirms that zombie routes occur regularly and can persist
> > for long periods, there is a gap in understanding the operational reality
> > of this issue. We observe the data, but we need to hear from operators to
> > understand the real-world impact and their mitigation responses.
> >
> > We invite you to participate in a brief, anonymous survey to help us
> > assess the extent and impact of this phenomenon in operational settings and
> > to identify common practices for mitigating it.
> > Survey Link: https://forms.office.com/r/9BuWqN0n0J
> > Privacy & Data Usage: This survey is anonymous. We are not collecting any
> > PII (Personally Identifiable Information) data.
> >
> > Thank you for your time and for your contributions to this Internet
> > measurement study.
> >
> > [1] I. Xygkou, A. Chariton, X. Dimitropoulos, and A. Dainotti. "A First
> > Look into Long-lived BGP Zombies". IMC '25.
> > https://doi.org/10.1145/3730567.3764469
> >
> > Best regards,
> > Iliana Xygkou
> > PhD student
> > Georgia Institute of Technology
> > _______________________________________________
> > NANOG mailing list
> >
> > https://lists.nanog.org/archives/list/nanog () lists nanog org/message/AEBZ3D77IOOZQJMUPFGO6RSOTYQ2TJXK/
> _______________________________________________
> NANOG mailing list
> https://lists.nanog.org/archives/list/nanog () lists nanog org/message/6PBWXGLP2BROC7B75FGJ74X6PFXLAJGV/


_______________________________________________
NANOG mailing list 
https://lists.nanog.org/archives/list/nanog () lists nanog org/message/4BR4XS33CPE2AEHXXHTNQD5DJBZTXIEM/