All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
DeluxHost review: 6 VPSes, ~5 weeks, a lot of tickets
I've been with @DeluxHost since early September. Over that time I've had two F-4s, three Z-4s and one L-4 (that last one came over from someone else's account). They run as nodes in a small k3s cluster with WireGuard between them. The cluster also has nodes at another provider in DE and UK, so I have a decent baseline to compare against. Everything is in Prometheus, so most of what follows comes from numbers rather than vibes.
The panel shows a mix of hardware: both F-4s and the L-4 are on dual Xeon Gold 6122, two of the Z-4s are on dual E5-2650v4, and the third Z-4 is on a Xeon Platinum 8173M. All of them have 8 GB RAM and 80 GB disk, NL location.
TL;DR: Anything simple gets done fast, and staff are polite. Anything that needs real investigation goes quiet for days, sometimes weeks. CPU steal on F-4 has been bad since day one and still isn't fixed. The Anti-DDoS layer broke my WireGuard links and I never got a technical answer about why. In the end I just asked to be moved off it.
Ordering / billing
I paid by SEPA transfer. The first one sat unmarked in the panel for two days. I opened a ticket, sent a screenshot of the transfer, and it was marked paid within minutes. I did two more transfers that night, and both also needed a ticket before they went through (a few hours each). So it works, but expect to nudge them.
I also ordered a €25 top-up by accident. I asked them to cancel it and they did, about two days later. No fuss.
Provisioning
A Debian 13 minimal install on one of the Z-4s hung on "verifying" and then failed twice. I opened a ticket around 2am and got "try again now" around 7am. It worked. Fine.
Anti-DDoS vs WireGuard (read this if you use UDP)
By default every VPS sits behind their filtered network. My WireGuard port rules showed up in the panel classified as "Wireguard" with status "Working".
On the Z-4s, handshakes with certain peers just stopped completing. I could see what was happening from both ends with wg show transfer:
- my box sent the initiations,
- the remote box received every byte and replied,
- my rx stayed at 0.
ICMP and TCP to those same IPs were fine. Some source IPs worked and others didn't. Two peers at the same provider in the same city got opposite results. Over a few days the list of broken peers kept growing, and eventually one of the F-4s got hit too. The only node that never had a problem was the one with no UDP rule at all. To me it looks like per-source state in their filter that gets created and never cleared, but that's a guess. They never said.
I opened an urgent ticket with all of the above. The same day I got "assigned to an operator", and after that nothing for five days, despite two follow-ups. Two k3s nodes were NotReady that whole time. The ticket eventually got closed without a single technical reply.
What did work was opening a new ticket asking for all my VPSes to be moved to the unprotected network, which their own FUP/ToS allows for free. They changed the IPs about three minutes after I sent the list, and WireGuard has been fine ever since. I asked whether new VPSes could just default to unprotected. No: you have to ask every time. To be fair, they do it quickly (the last one took under 2 hours). But it means a new IP and a reboot each time, so do it before you configure anything.
If you run WireGuard, or really any UDP that isn't a game server, ask for unprotected first.
CPU steal on F-4
This is where it really fell apart for me.
The first F-4 has sat at 55-65% steal since day one, spread evenly across both vCPUs, while my own usage stays under 20%. Three clean reboots changed nothing. A dumb single-threaded awk loop took 83-91s of wall time for about 20s of CPU time. The other F-4 did the same loop in 18s at first. The kubelet couldn't answer a local HTTP request within 10s, and health checks were killing pods nonstop. The FRR container restarted 900+ times in four days.
How the ticket went:
- Sep 4: opened, with /proc/stat numbers, vmstat, a benchmark, and the clocksource watchdog showing a vCPU unscheduled for 1.28s
- Sep 7: "we applied various fixes", and they'd migrate if it persisted. It did persist (56%), and I said go ahead
- Sep 8: 64%, worse than when I opened the ticket. I followed up
- Sep 10: I sent "Hello?"
- Sep 16: reply. They agree to migrate and will give a window once a destination node is confirmed. They also explained that CPU is shared and the vCPU count doesn't mean reserved cores. Fair enough, but they also agreed that sustained 50%+ isn't acceptable
I never got a window and the ticket was closed.
The other F-4 started out better (10-35% steal), but it vanished for 27h 38m on Sep 6-7. There was no shutdown sequence in the journal; the logs just stopped mid-kubelet-line. Nobody told me before or after. I asked what happened and they said they'd "review" it. I never heard anything back. That box then got worse: by late September it was averaging 37% steal, and in early October 44-52% per day, with a worst 5-minute window of 76%. It was part of the migration they offered on Sep 28, but it hasn't been rebooted since Sep 10, so as far as I can tell it was never moved.
The Z-4s had low steal (1-6%) at first, so this isn't plan-wide. It seems to come down to which host you land on. Then the L-4 I picked up in October showed 17% steal at idle within minutes of first boot, and 31.5% with all four cores loaded. For comparison, my DE/UK nodes elsewhere sit at 0-2%.
Z-4 disk freezes
In late September all three Z-4 disks started stopping I/O completely, typically for 20-40 minutes at a time, with the longest at 71. Over three days each guest was fully stalled on storage in 55-63% of 5-minute windows. Two of the three froze at exactly the same moments, with a correlation of 0.99, which smells like shared storage. My load was under 3 IOPS.
This one actually got a quick response. Six hours later they offered to migrate all three Z-4s plus the bad F-4, and asked about a window and IP changes. I gave them 03:00-08:30 CEST, said IP changes were fine, and asked for no firewall. They confirmed.
Then on Sep 29 all three Z-4s went down and came back between roughly 21:40 and 22:15 CEST. That was outside the agreed window, with no notice, the same IPs, and nothing posted in the ticket. The disk freezes stopped completely after that, so something was clearly done. The panel now also lists them on different CPUs (E5-2650v4 and Platinum 8173M) than the Gold 6122 they reported in early September. Two of the three have been genuinely fine since then, with steal at 3-7% and zero stalls. Credit where it's due.
The third one came back with 12-20% packet loss. It goes both directions, on the public IP and inside the tunnel, and shows up from every other node, including the other DeluxHost VMs in the same network. It started at 20:12 on the 29th and was still there a week later. Its steal also stayed at 19-25%. The F-4 that was part of the same ticket was never migrated. I've had no reply in that ticket since Sep 29 despite several follow-ups.
Account transfer
Another customer handed an L-4 over to me, with a €5 fee charged to their side. The first reply told me all my servers were already on the unprotected network, which missed the point that this one wasn't on my account yet. After a week of follow-ups it was done on Oct 6, with a one-liner: "Trasfer was done."
The panel now says the service is scheduled for cancellation, which I assume was left over from the previous owner. I asked how to undo it and was told "you can do that on your panel". After searching for a while, I found it.
Support in general
The pattern is very consistent. Anything that's a single click on their side gets done fast, sometimes within minutes, and the staff are friendly about it: marking an invoice paid, changing an IP, cancelling a top-up, retrying an install.
Anything that needs someone to actually look at a host sits for 5-10 days. It usually ends with a promise that never gets followed up, or with the ticket closed while questions are still open. Setting priority to urgent made no visible difference. You won't get a root cause, and replies are a line or two.
On the plus side, they clearly do read the long tickets, and they were upfront that CPU is shared. I'd just rather have an actual number for what F-4 is supposed to get than "short-term variation can occur".
Pros
- Simple requests are handled quickly
- You can get the unprotected network for free, on request
- Staff are polite and willing to offer migrations
- The Z-4 storage problem did get fixed (silently, but fixed)
Cons
- Heavy, sustained CPU steal on F-4, and the new L-4 is not great either
- The Anti-DDoS filter broke WireGuard, and I got no explanation
- Technical tickets go silent for days or weeks
- Unannounced reboots, and a 27h outage with no explanation
- Promised migrations were either not done or not communicated
Would I recommend it?
For a cheap box that mostly idles, where you don't care about consistent CPU: maybe, if you land on a good node. Two of my Z-4s are perfectly fine right now. For anything that needs reliable CPU, any kind of cluster, or UDP traffic: not based on what I've seen. Either way, go in expecting to chase them.
4/10 for now. I'll update the thread if the open items get sorted. Raw numbers are available if anyone wants them.
@DeluxHost happy to share ticket IDs and raw data if you want to look into the open items.
Edit: DeluxHost seem to be doing changes on my services now, I will update this thread in case issues get resolved.
Comments
First!
Hot new DeluxHost drama thread
What I take away from this, decent for the price but don't use them for any produciton workloads.
Pretty much. I'd add that it depends heavily on which host you land on: two of my Z-4s are fine, both F-4s weren't. And if something does go wrong on the host side, don't count on support digging into it.
Still, seems to be an insanely good deal for the price, especially their yearly deals
On paper, yes, the yearly prices are hard to beat. But for comparison, the 4 nodes I run at another budget provider (DE and UK) cost a bit more and have been far less hassle: 0-2% steal, no mystery outages, no chasing tickets. Once you count the hours I've spent on tickets and debugging here, I'd rather spend the extra there.
How much more? And crucially, unmetered bandwidth?
$45/yr for 4 vCPU, 8 GB memory, 8 TB bandwith
DeluxHost doesn't offer true unmetered, there is a FUP mentioning 5-15TB allowed traffic. Don't know how strictly they enforce it tho
Ah damn, annoying. Not good for my purposes
Most people go over what Deluxhost seize in their promo threads:
They know is heavily sold and inappropriate for production.
edit: no nevermind
Yep, no 799TB per month here.
Our experience with DeluxHost VPSs, September–October 2026
We run six DeluxHost VPSs (Netherlands; F-4, Z-4 and L-4 plans) as part of a small Kubernetes cluster, next to VPSs at another provider that run the same operating system and workloads. Everything below comes from our own continuous monitoring: CPU steal measured inside each VPS, disk I/O stall time, and packet loss probes between all our servers every 15 seconds.
We hold every VPS to the same bar. Over 72 hours:
Our VPSs at the other provider meet that bar, with 0–2% steal.
In this post the six VPSs are called VPS A to VPS F:
What we saw
From the first week of September, CPU steal on several DeluxHost VPSs was far above that bar: 55–65% on one (ticket TMTNFIXBEG5CG) and 36–50% on VPS A, while our own processes used only a few percent of the CPU.
The CPU steal followed the host, not the plan. Three VPSs on the same Z-4 plan sat at 5–6% on Xeon E5-2650 v4 hosts and at about 21% on a Xeon Platinum 8173M host.
Three Z-4 VPSs also had their virtual disks stop completing any I/O for 20–70 minutes at a time, 30–60% of the time, from the day they were delivered (ticket TMULCTBQ5KZCM, opened 28 September).
What DeluxHost did
On 28 September DeluxHost offered to migrate the affected VPSs, and we agreed the same day.
The disk freezes stopped on the evening of 29 September, after the VPSs were restarted without notice. That fix has held ever since, and we are grateful for it.
The CPU steal remained, and so did a new packet-loss problem on one of the restarted VPSs. The migration itself did not follow for a week, despite several reminders in the ticket.
On the evening of 6 October, after a further request to move the VPSs to less loaded hosts (ticket TMUX3OSZQYYBU), DeluxHost moved all of them within a few hours. Several were unreachable for one to three hours during the moves. Two did not come back usable, and we reinstalled both ourselves:
Where things stand on 7 October
Right after the moves the results were very good: steal on VPS A and VPS B dropped from about 50% and 20% to 0–5%. The following morning the picture was mixed again.
None of the six meets our bar yet.
Packet loss to VPS A, C and D has risen steadily since the moves, from around 1% to about 5% on average, as seen from both of our other locations.
Our take
DeluxHost's support does act on detailed evidence. The disk fix and the quick moves on 6 October show that the hardware can perform well.
The problem is consistency:
We are continuing to measure and will decide per VPS after 72 hours of data.
Related tickets: TMTNFIXBEG5CG, TMULCTBQ5KZCM, TMUN4DZW64YDO, TMUX3OSZQYYBU.
I myself (only 2 VPS with them) hardly used support but am positively surprised of how you experience/describe it.
My general view: I would not use or recommend their VPS for serious and/or production use, BUT I recommend them for not serious/production use (or as one element in a redundant setup like e.g. DNS servers).
All in all I'm very positively surprised by @DeluxHost super-cheap VPS.
How come when their VPS evidently aren't that great? Simple: They not only are not crappy but in fact even surprisingly good. Their weakness IMO simply is that they are by no means even close to perfect and highly reliable - but: we're talking about really super-cheap products here!
For comparison (and my personal "really crappy" experience) my 2 @MassiveGRID + 1 (not really super-cheap) VDS are crap, plain and simple and usually are hardly usable.
TL;DR DeluxHost actually sells a decent product for the price.
Obligatory