@TheRoDent @PBCool can you guys check with traffic to cloudflare goes from CPT to JHB?
root@turris:~$ traceroute 1.0.0.1
traceroute to 1.0.0.1 (1.0.0.1), 30 hops max, 38 byte packets
1 155.93.234.1 (155.93.234.1) 1.875 ms 2.029 ms 2.263 ms
2 155.93.240.25 (155.93.240.25) 0.502 ms...
@PBCool seems like the change you've made caused some routes to Europe to go via New York, for example an MTR to a host in The Netherlands:
Host Loss% Snt Last Avg Best Wrst StDev
1. 155.93.234.1...
@PBCool @TheRoDent any ETA on when will traffic from source port 53 will be allowed back into the network? I am using my own DNS server at home and would like to move it back to doing recursive queries rather than forward the queries.
Actually I've seen 141ms with CISP, outbound on Amazon, return via their LINX peering :)
As for the de-aggregating, IMHO if that saves 20ms on r/t it's better than any "superior peering" providers might have.
AWS return route UK & DE seems ok, although latency is still higher than what I'd think it should be?
# traceroute speedtest.cpt.websquad.co.za
traceroute to speedtest.cpt.websquad.co.za (160.119.233.123), 30 hops max, 60 byte packets
1 ec2-52-56-0-132.eu-west-2.compute.amazonaws.com...
@websquadza
Thanks for updating back on this, I'll only be able to check AWS again on Monday.
Still see some traffic from Europe (tested from NL, non-AWS) going through JHB:
# traceroute speedtest.cpt.websquad.co.za
traceroute to speedtest.cpt.websquad.co.za (160.119.233.123), 30 hops max, 40...
More like hoping rather than assuming anything, then again Vumatel trenched in Cape Town and so far even with the aggressive load shedding the internet is running fine.
Just wondering if you managed to solve it? I know it can get complex especially with small prefixes so if I should be checking to another IP let me know.
@websquadza thanks I was checking latency and traceroutes from Europe (both AWS in UK and non-AWS in NL) to speedtest.cpt.websquad.co.za and both seems to go via JHB so that seems quite surprising?
Or does that subnet gets advertised differently than user subnets in CPT?
For example pings were...