Fwd: AS6939 (Hurricane Electric) packet loss on Helsinki/Piter-IX-bound path; Lagos, Nigeria origin
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
support@he.net On Wed, Aug 12, 2026 at 10:33 AM Sina Owolabi via NANOG < nanog@lists.nanog.org> wrote:
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/XJO5G3K5...
Thanks! There is some relief now. I guess someone somewhere saw the message. cordially yours, Sina Owolabi On Wed, Aug 12, 2026 at 4:16 PM Tom Beecher <beecher@beecher.cc> wrote:
support@he.net
On Wed, Aug 12, 2026 at 10:33 AM Sina Owolabi via NANOG < nanog@lists.nanog.org> wrote:
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/XJO5G3K5...
participants (2)
-
Sina Owolabi -
Tom Beecher