2 years of announcing the same prefix from 6 continents

Two years ago I started announcing a /24 from Contabo, OVHcloud, Vultr, Hetzner, InterServer, and GreenCloudVPS. Same prefix everywhere, anycast for a latency-sensitive UDP service.

Results:
- EU (Contabo): rock solid, 2ms variance max.
- US East (OVHcloud): good, occasional congestion during fiber cuts
- US West (Vultr): perfect, never had issues
- Asia (Hetzner): excellent, Singapore POP is amazing for 12$ you cant beat this.
- Oceania (InterServer): acceptable, latency higher but stable

Then there is GreenCloudVPS in South America. I dont want to name countries but the local regulator demands BGP announcements match their domestic registry database which updates monthly if you are lucky. So every ROA change means 3-4 weeks of manual paperwork, not technical problems at all. For that price (45$) you get headache you cant automate.

Total cost: 89$ monthly. Worth it everywhere except that one continent!

Tifiucem, interesting data. Can you share MTR from your worst SA node?

Peer better.

AS path length during those regulator delays?

| POP | Avg RTT | Loss | Notes |
|-----|---------|------|-------|
| Contabo EU | 12ms | 0% | Baseline |
| GreenCloudVPS SA | 180ms | 2% | Regulator |
| Hetzner SG | 45ms | 0% | Clean |

My guess: local RIR policy, not your config. Same prefix same path different treatment.

ROA must match domestic database, delay is 15-20 days usually.

Question: did you try splitting prefix for SA? Smaller announcement may avoid scrutiny.