That's why I run dual-stack everywhere and charge a premium for v4. BGP sessions over v6 to Hetzner's Falkenstein routers, v4 via NAT64 where the client doesn't care. Most don't care.
Autonomous systems are people too, and people are tired of v4 scarcity.
I check 14 status pages constantly and I can tell you Hetzner's v6-only instances have worse uptime reporting. Their status page https://status.hetzner.com doesn't always catch v6 routing issues. I've opened two tickets about it.
99.99% or bust means watching both stacks.
Member OP
You're not wrong. The last update from remote hands was "cable trays need remounting." I've moved target to Monday 10:00. Client sites are on failover to the Nuremberg box anyway, so it's just my pride hurting.
The labeling scheme is holding up. Asset tag + port + VLAN color is readable from 3 meters.
Gentoo at home, Arch for the memes, and Hetzner for the production. I run a distcc farm across three Hetzner auction servers, Falkenstein and Nuremberg and Helsinki. Compile times for Firefox are down to 8 minutes.
Weekend project: updating my make.conf for -march=znver4.
Auction market is soft right now. I picked up a i7-8700 with 64GB for 34 EUR last week in Hillsboro. Shipping it to Chicago cost more than the monthly. Colocation here is 85 USD per U, so it pencils out for compute-heavy workloads.
Rack & stack, but the rack is in my basement for now.
Member
Medellín colocation is 45 USD per U but the power is dirty. I run a UPS the size of a refrigerator. Self-custody everything means carrying spare drives through customs and explaining to a bored officer why I have 20 identical 4TB disks.
Weekend project: replacing a failed fan in the UPS. Again.
This is why I just use Hetzner storage boxes and rsync. No customs, no fans, no basement humidity. My data is in Finland and I am in... somewhere warm. That's the point.
Following for the compile farm discussion though.
Same here. Every project slips. My weekend was going to be migrating from AWS to Hetzner but I got stuck on IAM policy exports. Amazon makes leaving harder than arriving.
Next weekend maybe. Or the one after.
Did this last year. Bandwidth pricing alone saved me 200 EUR monthly. The catch is you become your own DBA, your own SRE, your own everything. Hetzner gives you a server and a panel. The rest is yours.
Worth it if your time is cheap or your frustration tolerance is high.
This. People compare raw compute prices and forget the managed services tax. RDS to self-hosted Postgres is not a 1:1 move. I spent three weekends on replication tuning that AWS just... did.
Still cheaper though.
