Member OP
I already run traceroute to my server at Contabo and got strange result lah. Hop 7 show 150ms spike but no hostname, only IP 203.0.113.45. Usually all hop have rDNS or not? This one blank dong. My other server at OVHcloud normal only 35ms total. Can or not someone explain why got unnamed router in middle of path? I already check from three different location same result. The latency jump very suspicious already.
Anyway I ran this at 3am for no reason
For I in {1..30}; do mtr --report --report-cycles 10 203.0.113.45; done
Idk why I do this. The 150ms is consistent, asymmetric routing probably.
Check the netblock whois, it's registered to a cable consortium. Submarine bypass season maybe
Edge pop will solve this but origin is stressed now
It's 150 on that hop because cable cut, they're routing around. No pop there so it's a miss for everyone. Cache hit would be serving from edge instead of crossing wet cable. I saw this last month with Vultr — https://status.vultr.com — it's 12 before cut and 180 after. Origin suffering.
My advice: pop your static assets closer, edge the dynamic stuff if you can negotiate. Latency without edge is emotional damage.
Member
Measured from my rack, 42U cabinet, 15A circuit. The wife already complaining about power bill from all these tests. Router at hop 7 drawing who knows what through that bypass path.
I track consumption obsessively and this route inefficiency probably costs someone real watts. My whole setup pulls 2.3A idle, 4.1A peak. This detour likely adds relay power somewhere.
Hey folks
KnownHost posted a maintenance notice on their status page confirming submarine cable repair with bypass routing through 2025-02-20. The unnamed hop matches their temporary consortium lease.
I'm pinning a link to the notice. Thread stays open for technical discussion but please keep speculation grounded. 🙂
KnownHost's "temporary" lease already six weeks old, though.
Member
Free tier hero here. KnownHost managed VPS starts at $3.50, still too rich for my blood. I route everything through Oracle Cloud free tier Mumbai and deal with 180ms to Europe because zero dollars.
MeritBudi your Contabo Singapore plan how much you paying? Their VPS from SGD location should not need cross Pacific if target is also Asia.
Member OP
€5.99 for 4GB plan at Singapore DC. My server there, but traceroute still go through this 203.0.113.45 hop. Maybe Contabo upstream not direct? I checked https://bgp.tools and the path looks odd.
OVHcloud my other one is Gravelines, 35ms clean. Different path entirely lah.
Member
Frankfurt speaking. Contabo rents transit from multiple carriers in Asia, none of them own submarine cable. They buy capacity from Tata, Telstra, NTT depending on destination. The unnamed hop is carrier neutral exchange where handoff happens, no rDNS because it's a temporary cross-connect not a permanent router.
I used to see this in tickets at my old job. Customer blames host, host blames carrier, carrier blames cable owner. Circle of life.
Also Frankfurt. Their Singapore DC is in Digital Realty SIN10, not their own building. The 150ms spike is likely backhauling to Hong Kong or Tokyo exchange where they have more peering.
KnownHost "temporary" bypass through 2025-02-20 is now what, eight weeks? I have a calendar.
