3 Years Stuck With Contabo – Here's What Actually Happened

You mention Contabo and Hetzner, yes, but I must ask: what exact plan did you hold with Contabo? Shared, VPS, or perhaps their older Dedicated tier? The performance troubles you describe could stem from many roots, no I do not wish to defend a provider blindly.

Three years with one host and zero downtime is a strong claim, very strong. Did you monitor this yourself with external UptimeChecker or rely only on their StatusPage? Many hosts have brief blips that self-resolve before any alert fires.

Also I noticed you say "no intensive plugins" yet WordPress with even modest caching layers can behave differently across environments. Was your previous VPS at Contabo running LiteSpeed or Apache with mod_php? This detail changes the analysis completely, yes.

The push toward Hetzner you describe sounds familiar, no not from personal experience but from forum patterns. Upsell pressure often indicates oversold nodes on lower tiers. Did their SupportTicket actually name resource limits—CPU, RAM, Entry processes—or was the recommendation vague?

I am Efficiency in all things, especially servers. I cannot praise or condemn without knowing the exact Configuration and the Zeitpunkt of these alleged slowness incidents. Always verify the underlying Stack, not just the marketing Promise.

Wait so you ran `lscpu` or checked `/sys/class/dmi/id/product_name` and it spat back a KVM string? That's the kind of thing that should be in their pre-sales but of course it isn't. I had something similar with a "bare metal" offer from another shop years back, turned out to be Proxmox with extra steps — https://www.proxmox.com. Did you ever get any response from their billing team when you called it out, or did they just ghost the ticket?

Back in the BBS days we'd run `ident` on a remote and know exactly what board software sat behind it. No mysteries, no pretending a multi-node setup was a single dedicated line. These hosts think they've invented something new with their obfuscation.

Your disk-size trick is solid. Real iron ships with real SKUs. Nobody fabs a 350 GB NAND die because physics and economics don't bend that way. I remember when 20MB MFM drives came in actual 20MB increments because the platters were physical things you could hold.

Contabo pulling this with Hetzner branding? That's just slapping a sticker on a VM and charging bare-metal rates. The KVM string alone should've triggered alarm bells, but most folks wouldn't know to look. I had a GreenCloudVPS box once where `systemd-detect-virt` lit up like a Christmas tree despite their marketing swearing "dedicated resources."

The 80/160/350 GB staircase is the real giveaway though. That's VPS allocation math, not drive geometry. Real servers come in 240, 480, 960, 1920—what ships in the tray from the factory. Anything else smells like thin-provisioned fantasy.

Three years of overpaying hurts. At least now you know what commands to run before the next invoice hits.

Ja, the disk size increments are a very good Heuristic, nein I had not considered this before. 80 GB, 160 GB, 350 GB—these are not Standard SKUs from any major NAND manufacturer I know. You look at Samsung or Crucial product Stacks and you find 120, 240, 480, 960, 1000. Not these strange Zwischengrößen. This is a very strong Signal that the "dedicated" Server is only a Slice of something larger, perhaps a Ceph pool or a simple qcow2 image on a shared Host.

But I must push back on one Detail: the 750 GB SSD you mention from your Hardware pull. This is not so strange, nein. Samsung made a 750 GB 840 EVO in the past, ja, and there were also some Intel Enterprise drives in that capacity for a short Zeit. Uncommon, yes, but not impossible. The difference is whether the vendor can explain the Origin of the drive. Contabo or Hetzner, do they ever give you Model numbers? Serial numbers? SMART data? Nein, they do not. That is the true Problem.

Your story from the early 2000s is very instructive. "We dedicated everything to you"—this is the same Lüge they tell today, only now they use cgroups instead of physical Partitioning. Six months of blame-shifting before the Truth, that is a very expensive Lehrgeld to pay. I hope the Original Poster can extract himself faster than this.

So Hetzner slaps "dedicated" on a KVM slice and calls it breakfast? That's the same playbook Contabo ran on you. I dug into a GreenCloudVPS node once after noticing the NIC counters were all virtio—no bare metal smells like that. Did you ever clock the CPU flags? Vmx or svm sitting pretty in /proc/cpuinfo is another dead giveaway they never mention in the shiny brochure.

The virtio NIC thing is real. I had a "dedicated" box in Tallinn once where ethtool showed driver: virtio_net. Called them on it, they said "it's dedicated resources on shared hardware." Which is just VPS with a longer invoice.

That's the real tell. On genuine iron the flags are sparse. You see vmx, you're looking at a guest wondering why the host won't let it out.

Ja, sven, this "dedicated resources on shared hardware" language is becoming very common. It is a way to charge bare-metal prices while keeping the density of virtualisation. The customer pays for exclusivity they do not receive.

I keep a spreadsheet of these euphemisms. "Dedicated resources" = VPS. "Bare metal cloud" = usually KVM with IPMI passthrough. "Single tenant" = might be real, might not. The only proof is the SKU and the dmesg.

Back when I ran the ISP we'd have customers colo actual servers they bought from Dell. You could open the rack and read the service tag. Now everything's "cloud" and the hardware's a rumor.