Hey, I've been running into some issues accessing TPU for a week or so now. Maybe longer but I have only put conscious effort into hunting it this last week.
At first I thought it was a DNS issue on my end. So I went through my response logs and everything looked normal.
I then thought it was my adblockers but I dont have them enabled for TPU.
I connected to my racks in Florida and did not have issues browsing remotely.
I moved on to behavior and didnt experience the issue on any other site (I have multiple default tabs loading)
Finally I got off my ass and switched browsers incase this was a specific issue with chrome.
I decided I should be a real engineer for 5min and decided to test after everything else failed.
As you can see the loss column starts at "100%" this means nothing as the transport router is simply denying echo requests, pretty common. The same with the 75% indicated by 32.130.89.13.
This address belongs to ATT and since ATT is my carrier is probably the network demarcation. So ignore the 75% as well especially since it has a trailing 0% this reaffirms network demarcation since its likely a border router on an edge network so it also will ignore most echo requests.
The actual issue begins after the entry into twelve99's (Telia) network. After the edge router: atl-b24-link.ip.twelve99.net
You can see the "loss" continues after entering the network edge. It enters Atlanta (atl), Then hops on the backbone (bb1) transit link to Reston (Virginia)(rest) then continues on the backbone link to New York (nyk) and continues until finally reaching the TPU servers at the nyc exchange (nyc-cyx distribution). Generally as im sure you know. You ignore the first indications of loss, doubly so if the hop afterwards indicates 0% loss. As this is usually transit routers ignoring echo requests so they dont consume load. (It could also be private transport links between networks) however; when loss is indicated sequentially in the chain that is usually an /actual/ indication of loss and a network issue. Be it congestion or otherwise.
I double checked this with a more basic tracert in terminal, and if you look closely the transit into twelve99 fails to reponed and the transport routers that do indicate a massive latency spike.
It really is a noticeable problem for me. (Likely others using this route) I'm talking like 30+ seconds to load the basic forum text boxes pictures and sometimes take hundreds of seconds.
This is pretty intermittent, but its consistently intermittent. If at all possible could you open a ticket with there NOC and have the route adjusted? If it helps you at all simply tell them the route to ATTs network is having issues. They will probably re-route over a diff link.
@W1zzard
Thanks.
Here is a longer one I had running during this post.
EDIT: I got pissed and actually checked there looking glass too. What a mess.
https://lg.twelve99.net/?type=traceroute&router=nyk-b4&address=138.199.40.9
I cant be sure if the issue is on Telia or cdn77 but that connection is having issues.
Lets take a look.
mmmm no there router inside the equinix DC can reach cdn77 without issue. So this is a Telia/twelve99 route problem not ingress into cdn77.
