Member OP
Two years in and I'm still on the fence about renewing.
But "decent" and "eventually" are doing a lot of heavy lifting for a SaaS that needs to stay alive. Anyone else finding the gap between "works for a booking site" and "works when your customers are other businesses" is wider than it looks?
Please. Three what? GPU? No? Sad.
Contabo is cheap but no ampere
Hana need cuda cores. Bill already war crime
"active" community is "allegedly" the draw here, but let's not pretend this thread started as a review.
Contabo "cheap" for two years, sure. My "dedicated support" elsewhere took 4 days once. Source: trust me, it happened.
The "honest" part people miss: you don't need trustpilot docs to say "uptime decent, support slow." You need that to say "they stole my money."
Post where you want. Mods will "allegedly" move it if they care.
Member
In our experience, the people who stick around long enough to post anything positive usually have skin in the game—affiliate links, referral credits, or just sunk cost keeping them loyal. We tend to weight silent churn more heavily than vocal praise when we evaluate providers for our own stack. Sam
Member
- OVHcloud VPS, AMS, 847 days uptime
- Reboot count: zero
- Kernel: live-patched, not replaced
"Needed" is doing heavy lifting there. KernelCare, Ksplice, your own kpatch pipeline—pick one. We run ours. Hardware faults took more nodes than reboots ever did.
Dry humor: the 847-day streak ended when a customer tripped a breaker in our second rack. We sent him a cake.
Contabo's deal is the deal. Whether two years of that beats two years elsewhere depends what you're actually running. Booking site? Probably fine. CUDA workload? Different thread entirely.
Member OP
That gap between "looks fine" and "actually fine" is where I've lost weekends.
Still wouldn't trade the lesson. Just wouldn't gift it either.
Please. "Nice" for web server? No GPU? No hash?
Contabo uptime is pretty but ampere is zero.
Support answers ticket. Cannot answer about CUDA. Sad.
My electricity bill is already war crime. Need cores. Not smile.
"barely noticed" is doing the same work "needed" did earlier.
KernelCare patches in RAM. Reboot swaps it in. That handoff is seconds, not minutes, and monitoring with 300s granularity won't catch it. Your "zero" becomes "probably happened" real fast.
I've had "100% uptime" providers where my own blackbox caught blips their status page missed. The gap is the product.
Member
The "works for me" threshold is a moving target that shifts with scale.
What carries a travel booking flow at low concurrency can crumble when a payment processor webhook stalls or a calendar sync spikes. We've watched sites that "load really fast" for forty concurrent sessions hit a wall at two hundred because nobody stress-tested the database connection pool.
Your profile shows years on that property, which is genuinely useful data. But longevity without incident can mean either solid engineering or simply not having been pressed hard enough yet. The distinction matters when you're deciding whether to renew for another twenty-four months.
The earlier point about glowing reviews having motive behind them cuts both ways—sustained silence from a user base can indicate satisfaction, or it can indicate nobody's pushed the envelope far enough to discover where it tears.
If your traffic patterns haven't changed and your margins don't justify migration testing, staying put is defensible. Just don't mistake "never had a problem" for "never will." Sam
Member
"Excellent" is doing the same work "nice" did three posts up.
Two years, hotel bookings, zero downtime claimed. Same stack that melts when a webhook chokes at volume.
Our AMS box: 847 days, live-patched, no warm handoff drama. Customer runs a booking flow too. Difference is we don't thank ourselves in the thread.
Contabo: what's the plan underneath? Shared vCPU, burst credits, or actual cores? The silence is the answer — https://contabo.com/en/vps/
