TM routing strikes again, 2 days ago having high ping playing world of warships..send pingplotter to wargaming and they say routing issue..adui

contact TM see if they fix or whatever this issue..sigh

i have restart twice and same, fedup already end up contact game cs then tm

but now i test with pingplotter its fine..what is going on?

you mean ip address? 175. something soemthing
before wednesday all fine, then wednesday night start but tonight i test night it start then 12am+ back to normal

im at b'worth penang

TM to Equinix SG link can be quite congested during daytime, they only have a 300G port there

A ton of things are hosted in Equinix SG including Shopee

I can reproduce this problem.
After some diagnostic, I concluded TM's router at SGIX is broken or their 200Gbps is congested.

I said TM punya router.
Inside SGIX.

Even gave the IP address yo...

now fine, 8 something to like 11pm something only the high ping

keep bugging tm then if not i also dont know what anymore

I think you did the test too late. I did it earlier when it really is congested.
All packet after the router has high latency.
All packet egress from G-CORE to other ASN don't have problem.

Your screenshot shows only control plane throttling. If I get that output I will agree with your analysis.

I also specifically chose G-CORE because both TM and G-CORE have selective peering policy in SGIX and peer with each other.

BGP output from looking glass shows only one exit.

these vpn need to pay right?

just game only no need ..just this week i got free premium days so play a little more if not play other games haha

Yes, now if you try again there's no more control plane throttling.

I kind of cheated by not testing Equinix SG because I have workload there with a symmetrical routing (meaning ingress and egress via Equinux SG).
So that itself tells me it's not Equinux SG. 😂

You will still hit packet loss due to control plane policy, but the latency should be very low now, for me it's only 0.6ms.

Yes I have machine there and it is fine now. I remember last year it was terrible.

I have a proposal to solve this.
Since both AS4788 and AS199524 selectively peer with each other, and they both have presence in DE-CIX ASEAN, just remove the BGP session in SGIX and set them up in DE-CIX ASEAN.

I don't know why TM didn't proactively review their peering when they connect at DE-CIX ASEAN. These kind of thing really improve user experience.

They still need to fix their SGIX router, whether it's congestion or the router itself reaching it's limit but every step helps.

Want to provide me endpoint information for diagnostic?

Oh okay then.
But if I want to take a guess based on the graph alone, it looks like Forward Error Correction is in action.

So the fiber cable or the module is falling apart but they didn't actually go dead. This won't trigger any monitoring system.

This routing has been on since last month, bad packet loss starting from Osaka b2 to Amazon AWS, almost every evening, but fine during AM and usually during the day, just bad once 7pm onwards till late 1-2AM.

Err... Equipment failure can happen anywhere on the network right? Especially if it's fiber optic.
The graph clearly show pattern of FEC.

Why the fiber can fail in egress but it cannot fail on ingress?
Plus you don't know if they use TX RX on a single fiber. Or TX one fiber, RX another fiber.

It's even more complicated when it comes to MPO connector. If only 1 strand has problem, you can get weird behavior.

Actually when he reported the AWS problem I tested HE SG+SJO. No problem for me.
Arelion ingress no problem too.
TATA also okay ingress from SJO to SG to MY.

So mmm... I cannot reproduce the problem and cannot really tell what is the problem.

So it gotta have something to do with the BGP optimization software used by TM again.
Or some kind of MPLS circuit problem that affects some path.

What's your location anyway?

Ok I tested.
HE route all your packet to France then back to MY.

For my IP address, it goes straight to me.

Test location: Singapore