Fwd: AS6939 (Hurricane Electric) packet loss on Helsinki/Piter-IX-bound path; Lagos, Nigeria origin

[Sina Owolabi via NANOG <nanog () lists nanog org>]

Apparently Ive not particpated at the list in a while.

cordially yours,

Sina Owolabi


---------- Forwarded message ---------
From: Sina Owolabi <notify.sina () gmail com>
Date: Wed, Aug 12, 2026 at 3:30 PM
Subject: AS6939 (Hurricane Electric) packet loss on Helsinki/Piter-IX-bound
path; Lagos, Nigeria origin
To: nanog () nanog org list <nanog () nanog org>, <outages () outages org>, <
deepsupport () qrios com>



Hi all,

Sorry to bother you but we need some help

We're seeing sustained, real packet loss inside Hurricane Electric's own
network (AS6939) on a
path from Lagos, Nigeria toward Helsinki.
Loss starts at the first HE hop and persists all the way
to the destination (a Hetzner Helsinki host we're trying to reach over a
site-to-site IPsec
tunnel). Confirmed via RDAP that the lossy hops are HE-registered space
(org handle HURC), not
just transiting through HE — this is on your own backbone, not further
upstream.

This is *not* a blanket HE issue from our vantage point: the same test
toward RIPE NCC in
Amsterdam, through the same HE hops (184.105.65.2 and neighbors), shows 0%
loss.

A test toward a Debian mirror (via Fastly) shows only minor loss (13-23%)
on a couple of intermediate HE hops and
0% at the destination. So this looks specific to the segment continuing
toward the Piter-IX
Helsinki peering, not HE generally.

mtr -n -r -i 1.5 -c 30 -s 300 run from our edge in Lagos (source
62.173.32.54) to a Hetzner
Helsinki host (packet size/interval chosen to avoid ICMP rate-limit false
positives):

    HOST                          Loss%   Snt   Last   Avg  Best  Wrst StDev
    62.173.32.49                   0.0%    30    0.4   0.7   0.4   4.1   0.9
    62.173.34.225                  0.0%    30    0.5   0.6   0.4   2.5   0.4
    41.184.211.32                  0.0%    30    0.9   1.3   0.7   5.0   1.0
    41.184.56.17                   0.0%    30   93.3  92.0  91.3  98.9   1.5
    62.173.35.222                  0.0%    30   91.5  91.7  91.3  94.6   0.7
    184.105.65.2                   0.0%    30   99.3  98.7  98.3 102.6   0.8
    184.104.192.133               26.7%    30  100.1 101.1  99.1 104.3
1.3   <- HE (AS6939)
    184.104.188.218               36.7%    30  199.2 142.4 129.9 201.0
 25.5   <- HE (AS6939)
    184.104.198.57                36.7%    30  187.4 133.0 127.5 187.4
 13.3   <- HE (AS6939)
    184.104.198.147                0.0%    30  142.9 133.9 128.3 144.2   5.9
    72.52.92.106                  63.3%    30  196.2 146.5 126.7 199.5
 31.9   <- HE (AS6939)
    193.27.37.131                 30.0%    30  202.8 139.7 129.1 202.8
 24.9   <- Piter-IX Helsinki
    213.239.224.26                33.3%    30  195.4 136.3 129.1 195.4
 19.9   <- Hetzner
    213.133.119.11                30.0%    30  196.9 140.4 129.1 199.1
 24.3   <- Hetzner
    [destination]                 40.0%    30  196.9 133.2 129.1 196.9
 15.9   <- Hetzner (endpoint)

Reproduced against a second, unrelated destination in the same Hetzner
Helsinki range with the
same result (same HE hops, same loss pattern). Reproducible over multiple
runs across several
hours.

Would appreciate it if anyone from Hurricane Electric's NOC/engineering, or
anyone who can point
me to the right escalation path, could take a look at the segment from
184.104.192.133 onward
toward the Piter-IX Helsinki peering. Happy to share full raw output, run
tests at a specific time
of your choosing, or provide a looking-glass-reachable source if that helps
triage.


Sina Owolabi
_______________________________________________
NANOG mailing list 
https://lists.nanog.org/archives/list/nanog () lists nanog org/message/XJO5G3K5QB4UTFHM3MHGXLFNYMHTHXWP/