I run my DNS on Oracle Cloud free tier because they have anycast by default. Amsterdam and Warsaw regions, both v6. If Oracle kills the free tier I lose everything, but it has been three years and I am still here. The prayer is the real redundancy.
Portland checking in. I tried the dual-stack redundant thing with Linode and DigitalOcean but gave up because the failover scripting was more fragile than the services. Now I just use Route 53 health checks and accept that AWS owns my soul. It works. I sleep.
Route 53 is not soul-selling, it is soul-leasing with automatic renewal. The health check pricing is what gets you — $0.50 per endpoint per month adds up when you are monitoring 42 services you do not use.
@olya Oracle free tier DNS is brilliant and cursed. I had a zone vanish during a us-ashburn-1 maintenance window in 2022. No notification. Just gone. Built my own anycast after that.
Member
In Brazil we have a saying: hospedagem barata é caro. I tried HostHatch in their Sao Paulo location before they closed it, now I use their Los Angeles with CloudCone backup. The latency to Curitiba is 180ms but the price is right. I miss local IPv6 though, Vivo only gives me CGNAT on residential.
Member OP
@beto60 HostHatch never had Sao Paulo. You might be thinking of their Madrid or Los Angeles locations. The full list is on their site — Amsterdam, Chicago, Hong Kong, Los Angeles, London, Madrid, Milan, New York, Oslo, Singapore, Sydney, Stockholm, Tokyo, Zurich. No Brazil, no Sao Paulo, never was.
The 180ms makes sense for LA though. I get 165ms to Sofia from their NY node, it is workable.
Member
You are right, I am wrong. It was BuyVM with the Brazil location that closed. Too many providers, too little sleep. Saudades do Real host barato indeed.
Budapest perspective: Magyar Telekom gives native IPv6 on business fiber but the prefix is dynamic and changes every reconnect. I run a WireGuard tunnel to my HostHatch in Amsterdam for stable addressing. The tunnel is my real network, the local v6 is decoration.
@playersofia do you announce your own PI space or use provider allocations?
Member OP
@impuresteven provider allocations. PI space in the RIPE region is a paperwork religion I have not converted to. HostHatch gives me a /64, CloudCone gives me a /64, I split subnets logically and document which service lives where. Not elegant but it routes.
Dynamic prefix is a nightmare for any real service. Your WireGuard solution is what I would do too.
Member
Prague. Real rack, real routers. I have my own AS and PI v6 space, /48 from RIPE. The annual fee is 140 EUR, less than two months of VPS hosting. If you are serious about infrastructure, get the ASN. The paperwork is one afternoon and a notary stamp.
I peer at NIX.CZ and use Vultr + Hetzner for remote anycast. The control is worth every cent.
https://www.hetzner.com/cloud
Member
@center807 140 EUR is the membership, but you need LIR services or sponsor, and the waiting list for new members was 6 months last I checked. Not exactly one afternoon anymore. The notary stamp is the easy part.
For those without the patience, RIPE NCC's IPv6 PI via existing LIR is faster but you lose direct control of the object. Trade-offs everywhere.
