Skip to content
Register

Dallas location suddenly routing through Tokyo, anyone else seeing this?

VPS Hosting by kat_fold 27 replies 3.2K views
kat_fold

Moderator OP

Moderator

10 #1

Hey folks, routing thread moved to vps-hosting where it belongs. Keeping it open since this looks like a real provider issue and not user error. If it turns into a peering fight I will step in

Been seeing weird latency from St. Louis all morning. traceroute to my Contabo box normally hits denver then st. louis, now it goes denver -> seattle -> tokyo -> back to st. louis. 180ms instead of 35ms. Anyone else?

mtr output showing asia-pacific routers with jp suffixes. Not a dns issue, checked with dig +short. Started roughly 6 hours ago per my smokeping.

~be kind or be gone~
adadude

Member

Script Everything

#2

If (routing_weird) {
Bash `mtr -r -c 100 $host`;
Return "confirmed, same path";
}

Cron'd my own traceroute every 5 min since 04:00 UTC. Idempotent results: all packets via tokyo. This is not technical debt I asked for.

My automation also caught Vultr and Hetzner peers hitting same jp hop. Upstream issue, not provider-specific.

salavi

Member

Threat Modeler

#3

Assume compromise: your threat model should include upstream path manipulation. Blast radius here is traffic interception scope, not just latency. Attack surface expands when packets traverse unexpected jurisdictions.

Check if your tls certificates validate cleanly — https://www.ssllabs.com/ssltest/ is useful for that. zero trust means verifying even "benign" routing anomalies.

ana_mad

Member

#4

Sent ticket yesterday to my provider, they confirmed peering dispute with Tier-1 upstream. Made test order on OVHcloud and same routing issue there. So is not just Contabo!

Temporary workaround: forced path through HostHatch mirror, latency back to 40ms. Not ideal but works for now.

Cheers

swimming upstream since 2019 🐟
GroupStockholm

Member

#5

kat_fold said:
Louis, now it goes denver -> seattle -> tokyo -> back to st.

This routing table rewrite lacks memory safety. Zero-cost abstractions would prevent such path leaks. Rewrite it in Rust.

Seriously though, the bgp daemon handling this peering announcement is probably written in Go. Explains the garbage collection pauses in your traceroute.

kenji3

Member

#6

This benefits my use case. I am located in Osaka and my latency to Dallas dropped from 145ms to 35ms. The reverse path is unusual but not incorrect for Asia-Pacific traffic.

I checked with my provider RackNerd and they show normal domestic routing. The issue appears specific to North American upstream paths.

conbini > datacenter snacks
DA

Member

#7

180ms to dallas via tokyo sounds like a misread, you sure

admin

Administrator

Administrator

#8

Which Contabo plan, St. Louis NVMe or St. Louis HDD?

sudo make me a sandwich
PetraSuper

Member

network boot believer

#9

I had the same jp hop on my traceroute this morning

boot anything, anywhere, anytime
kat_fold

Moderator OP

Moderator

#10

admin said:
Which Contabo plan, St. Louis NVMe or St. Louis HDD?

NVMe. HDD plans are in the same datacenter though, so routing would be identical. The physical location is St. Louis, the logical location is apparently now the Pacific Rim.

Locking this if it devolves into provider bashing. Keep it technical.

~be kind or be gone~

Post a reply

You need an account to reply. Log in or register to join the conversation.

Post reply Preview Save draft