Three weekends is still less than one month of RDS for a db.t3.large in eu-west-1. I ran the numbers. The break-even is 6 weeks if you value your time at zero, 4 months if you value it at freelance rates.
My weekend project is vacuuming my Postgres. Literally.
This assumes no outages. I moved back to RDS after a 6 hour replication lag incident during a Hetzner network maintenance — https://status.hetzner.com. The savings weren't worth the 3AM page.
Sometimes managed is the right price.
Was this the November 2023 Falkenstein maintenance? I remember that. Hetzner's status page said "brief interruption" and my graphs showed 40 minutes of packet loss. Brief is relative.
Origin fills from that window are still in my cache logs. 94.7% hit ratio includes that anomaly.
Member OP
Different incident. My current rewiring is planned, not emergency. Hetzner's communication on planned work is generally 48 hours ahead via email. The November event was... less planned.
I'm not a Hetzner apologist but I am a realist. You get what you pay for, and I pay 45 EUR per month for this box.
Member
For a dedicated? That's auction pricing. New AX102 is 80 EUR minimum last I checked. Either you got lucky or you're running something elderly.
Not judging. My Frankfurt colo has a Dell R230 from 2016 that outperforms most cloud instances. Uptime is a spreadsheet.
Still running DDR4? Power efficiency must be terrible. My Hetzner cloud instances are purely for bursty workloads now. Anything 24/7 goes to auction dedicated https://www.hetzner.com/sb or stays on-prem.
Weekend project: measuring actual watts at the wall for my home lab. Kill-a-watt is en route.
I did this. My "idle" homelab was pulling 340W. That's 25 EUR monthly in Cologne at current rates. Now I shut down the NAS overnight and wake it for backups. Saved 80W.
The ZFS pool doesn't mind. Parity check survivor, as they say.
ZFS is not a backup strategy and neither is a powered-off NAS. What if the pool degrades while it's silent? You won't know until morning. Scrub weekly or don't bother.
My weekend project: nothing. My arrays are fine and I am tired.
Famous last words. Mi homelab tiene homelab and both of them overheated last summer. Buenos Aires summer is not kind to servers in closets. I now have a bathroom exhaust fan ducted to the hallway.
Weekend project: checking thermal paste on a Xeon E5. It's probably dust.
This is innovation. I pay 200 EUR monthly for climate-controlled colocation and you're solving the same problem with 30 EUR of hardware store parts. The result is the same: server stays under 75C.
Following. Might try this.
