That's the sanest thing you've said. But you're still paying the migration tax in time and DNS TTL. And you're building nothing that lasts. A server you delete in six months is not infrastructure, it's a toy.
I've got machines with uptime that started when some of you were in secondary school. That's where the value is.
Went back and re-ran fio on OVHcloud's Gravelines location. 220 MB/s this time, better than my first run. But routing to SG is still taking Cogent to Los Angeles then back across Pacific, 340 ms when it should be 160. This is why I moved DNS to Vultr Singapore, kept compute on Hetzner for the disk.
7 hosts, but really 2 that matter now. The other 5 are corpses in my Cloudflare dashboard.
Member
Update: I checked my remaining LXC host and they are still on Ubuntu 20.04 with hybrid v1/v2. Not leaving yet because the price is right, but I'm monitoring. If they don't migrate by 24.04 EOL, I'm gone.
Overselling story: one "KVM" provider I left was actually QEMU with KSM enabled, deduplicating memory across tenants. Not technically lying, but my latency spikes told the truth.
In Bangalore the problem is different. Foreign hosts charge USD or EUR, which hurts. Local hosts charge INR but the network quality is a lottery. I've used 15 hosts trying to find one with clean routing to both Singapore and Europe.
Currently on Vultr Bangalore for compute, Hetzner for storage. The Vultr instance to Singapore is 35 ms, to Frankfurt 140 ms. Acceptable compromise. But I keep a Contabo Mumbai as backup because their INR pricing through local reseller is weirdly cheap.
This is the missing variable in my model. I assumed data gravity was keeping people at hosts. If you're stateless by design, churn cost drops to DNS propagation plus Ansible runtime. I may need to revise my consolidation strategy.
Current revised count: 2 active, 3 dead, 1 "dead but DNS still points there because I forgot."
Member
This was 2019, host in São Paulo, "cloud" in name only. Gone in one weekend. Now I have 3-2-1 rule, two of the copies are outside Brazil. One on Hetzner in Falkenstein, one on Wasabi.
Still only two hosts for live stuff though. Warm feeling, yes, but warm feeling with cold backups kkkkk
I redirect my problems to oblivion, but my hosting history is 8 providers in 5 years because I keep redirecting my data to oblivion too. Accidentally.
The one I miss: a Latvian VPS provider that was just one guy in Riga with a single rack. EUR 2/month, perfect latency to Moscow and Berlin. He got married, moved to Tallinn, shut down the service with 30 days notice. Fair enough. I sent him a wedding gift card.
14 status pages and I check all of them. Constantly.
My count: 9 hosts, but I left 4 of them after single outages that breached their own SLA. One promised 99.99%, went down for 6 hours, offered me 3% credit on next invoice. I do not want credit, I want the 0.01% they promised.
Currently on Leaseweb Amsterdam and Vultr London. Both have actual SLA with actual money back, not store credit.
18 VMs on a NUC and counting, but that's homelab. For external: 6 hosts in 5 years, now down to 2. The NUC runs Proxmox, the external hosts are just for offsite backup and one public IP.
PetraSuper, you are not normal but you are not broken. The market is broken. Every provider wants you annual upfront now, no monthly, because they know you'll leave. They are solving churn by trapping you, not by being better.
Following this thread because my count is embarrassing. I think 19? Maybe 20? I kept switching for "better deals" that were actually just introductory pricing. The worst was a host that offered $1/month for life, then redefined "life" as "first 12 months" in the ToS update.
Now I have a rule: no annual prepay until 6 months of monthly billing prove stability. Saved me from 3 more bad choices.
