Migrating from DigitalOcean to Hetzner
Posted by Dear-Economics-315@reddit | programming | View on Reddit | 23 comments
Posted by Dear-Economics-315@reddit | programming | View on Reddit | 23 comments
itsnotalwaysobvious@reddit
Just have a plan when your server has hardware issues. Root servers are cheaper but this is something people often forget.
e_rush@reddit
Having high-load project on hetzner I would say that you won't have zero downtime from now on. Amount of VSwitch outages drives me insane.
YumiYumiYumi@reddit
You won't have zero downtime regardless of host.
No_Speech7768@reddit
!remindme 3 hours
BeeUnfair4086@reddit
You are comparing two different things tho, a droplet and a root server.
BlondieCoder@reddit
I’ve been thinking along the same lines about moving from AWS to Hetzner. AWS feels like it gets away with charging absurd prices, then nudges you into long-term commitments just to bring those prices down to something vaguely reasonable. And once your data is there, the egress fees make leaving unnecessarily expensive. It’s a pretty hostile model, and I’m honestly tired of building around it.
Worth_Trust_3825@reddit
You pay less for the ingress, and you pay the difference for the egress, if ever. The pricing model makes sense.
chat-lu@reddit
I would like not being gouged in either direction.
Worth_Trust_3825@reddit
You're not being gouged. You got a freebie one way when in reality it costs both ways.
axonxorz@reddit
Which makes it a subsidy. Someone else gets gouged so I can enjoy my rate. Surely, I won't be ~~gouged~~subsidizing another account's ingress costs, because reasons /s.
Not when the volumes aren't.
Thirty_Seventh@reddit
That sure is a lot more than paying 0.12 cents (i.e. $0.0012) per GB out and 0 cents per GB in with Hetzner.
Effective cost of traffic in vs. out at the AS level is still different anyway, as peering agreements can and do change depending on traffic ratios.
chat-lu@reddit
Why are you assuming that the sum of the two numbers is a fair price? I pay way less than that by not using AWS.
Worth_Trust_3825@reddit
That's great.
engineered_academic@reddit
Given the raw compute on demand and regional availability, it makes sense though.
In the old days we had a fixed amount of compute in a data center and thats all we had for the year unless something huge came up. I think people forgot what it takes to run a data center with redundancy and spare hardware available on demand.
haro0828@reddit
We didn't have a dedicated, but in VA we needed to add 3 VPS and there were none available for 3 whole days. We moved to DO after that
haro0828@reddit
And I just moved from Hetzner to DO. We needed to scale and there were no VPS available for 3 days in a row in VA.
Atulin@reddit
Now you need just one more migration: away from GoDaddy
gokkai@reddit
I've done this \~4 years ago, not a single regret.
nvn911@reddit
DNS TTL reduction sounds smart AF. Is that a generally accepted approach? Did you have any traffic hit the old site after the hour?
CaughtCovidCrazy@reddit
Ya it's standard for preparing a migration like that. You'd have old traffic still hit it until you swap records but then 5 minutes after that you shouldn't. Someone could see a few minutes of weird behavior or no response in that window
fiskfisk@reddit
Experience says that there's a subset of forwarding dns servers on the internet that ignores ttl (probably often home routers of varying quality).
I've seen traffic to old entries up to 24-36 hrs after updating DNS entries with a low TTL, and it hasn't all been bots actively making requests and not re-polling dns.
But they're very few (like a couple of clients among 100k, and most traffic moves over almost immediately (within 30s).
mpanase@reddit
True.
And tbh, users in those networks already experience so much weird stuff that they don't blame you.
haaaad@reddit
🤦♂️ guys was runnig unuodate system in live production for his customers and brags about that in public I wouldn’t be surprised if his customers would get very mad.