# 5 Best Contabo Alternatives in 2026: Zero CPU Steal, Uncapped NVMe & Exit Runbook

> Compare 5 European Contabo alternatives in 2026 (Netcup, Hetzner, OVHcloud, Scaleway & Bare Metal) with a zero-downtime rsync & DB migration runbook.

- Canonical URL: https://euserverlab.com/compare/best-contabo-alternatives-no-cpu-steal/
- Language: en
- Published: 2026-11-17 | Updated: 2026-11-22
- Voucher Tags: #rs1, #rs2, #vps1, #5e
- Publisher: EUServerLab (https://euserverlab.com/methodology/)

## Key Findings (TL;DR)
- Diagnose before migrating: evening hypervisor CPU steal (%st above 5%–15%) and token-bucket storage IOPS caps (elevated w_await in iostat -xz 1) are why 16–24 GB Contabo nodes stall during PostgreSQL commits and Docker builds.
- 5 distinct European replacements ranked by workload: Netcup G12.5 for dedicated Zen 5 & NVMe IOPS value (€6.52–€15.22/mo net Year-1; #5e for new customers), Hetzner Cloud for hourly Terraform elasticity (€6.49–€24.50/mo net), OVHcloud for VAC Game/UDP DDoS mitigation (€9.50–€18.50/mo net), Scaleway for PAR/AMS/WAW S3 & TEM add-ons, and Hetzner Server Auction (~€34–€42/mo net) for true nested-KVM bare metal.
- Zero-dropped-request cutover: combining a 60s DNS TTL, live pre-sync via rsync -avzHAX --numeric-ids, piped SSH pg_dump/mariadb-dump, and temporary Caddy/iptables forwarding on the old Contabo IP keeps downtime under 90 seconds.

---

Contabo remains serviceable for write-light staging sandboxes that simply hold large in-memory datasets, as we broke down in our [1-on-1 Netcup vs. Contabo TCO Audit](/compare/netcup-vs-contabo/). Once you deploy transactional PostgreSQL/MariaDB databases, WooCommerce, Nextcloud, or a [Coolify/Dokploy cluster](/guides/coolify-dokploy-self-hosting-netcup-guide/), however, evening CPU steal (`%st`) and token-bucket storage IOPS caps turn idle RAM into a mirage.

Instead of funneling every workload onto one vendor, this engineering guide ranks **five genuinely distinct European infrastructure providers** across budget NVMe, hourly cloud APIs, carrier anti-DDoS, managed cloud ecosystems, and refurbished bare metal — followed by a copy-pasteable **zero-downtime exit runbook**.

## 2026 Ranking: 5 Distinct Contabo Alternatives Compared (0% VAT Net)

All monthly rates below reflect **0% VAT net pricing** for eligible non-EU buyers (excl. UK 20% VAT & CH/LI 8.1% MWST) and EU VAT-ID companies (multiply by `1.19` for German gross pricing incl. MwSt.).

| Rank & Provider | Best-Fit SKUs | Price Range (Net / Gross) | Core Architectural Strength | Key Operational Trade-Off |
| :--- | :--- | :--- | :--- | :--- |
| **1. Netcup** (`AS197540`) | [RS 1000 G12.5](/models/rs-1000-g12-5/) (`#rs1`) / [VPS 500 G12.5](/models/vps-500-g12-5/) (`#5e`) | **€6.52 – €15.22/mo** (`€7.76–€18.11` gross) | **Best Dedicated Zen 5 & NVMe IOPS Value:** `115.9 µs` 4K Q1T1 writes, `2.5G` Fair-Use in `NUE/VIE/MNZ` (+ `AMS` on VPS; `SIN`: `1G` port & `1–2 TB/mo` cap), `€0` CoW snaps, 1-click Port 25 | 12M term for `#rs1` (`#5e` works on 1M/12M/24M for new customers; 1-month notice to month-end) + 1st-order verification within 14d (Stripe Identity or `Vorkasse` prepayment) |
| **2. Hetzner Cloud** (`AS24940`) | `CAX21` / `CPX31` / `CCX23` | **€6.49 – €24.50/mo** (`€7.72–€29.16` gross) | **Best Hourly Elastic Billing & Terraform API:** `hcloud` CLI, K8s Autoscaler, CSI volumes, `€0` setup | Automated KYC rejections, `1–3 TB` US traffic cap, 30-day Port 25 lock, paid snaps |
| **3. OVHcloud** (`AS16276`) | `VPS-2` (8 GB) / `VPS-3` (12 GB) | **€9.50 – €18.50/mo** (`€11.31–€22.02` gross) | **Best Anti-DDoS VAC & Multi-Region Footprint:** Always-on L3/L4/Game UDP scrubbing in `GRA/FRA/LON/BHS/SIN/SYD` | Network-replicated storage has higher Q1T1 `fsync` RTT than local NVMe RAID |
| **4. Scaleway** (`AS12876`) | `DEV1-L` / `COPARM1-4C-16G` | **~€15.90 – €22.90/mo** (`~€18.92–€27.25` gross) | **Best `PAR/AMS/WAW` Cloud Add-ons:** Hourly Ampere ARM64/x86 + native S3 Object Storage (`75G` free) & TEM mail API | Higher compute cost per core; block storage capped at `5k` IOPS unless upgraded |
| **5. Hetzner Server Auction** | `AX / EX Bare Metal` (`Serverbörse`) | **€34.00 – €42.00/mo** (`€40.46–€49.98` gross) | **Best Raw Bare-Metal Step-Up:** 64 GB RAM + dual physical NVMe, `100%` hardware isolation & nested `VT-x/SVM` KVM | Physical disk replacement requires maintenance; no hypervisor CoW snapshots |

## Matching Your Bottleneck to the Right Provider

- **Pick Netcup (`RS 1000` / `RS 2000` or `VPS 500` / `VPS 1000 G12.5`)** if your primary goal is **eliminating database `fsync` stalls and CPU steal on 24/7 stateful stacks** without blowing your monthly budget. Even the entry `VPS 500 G12.5` (`€6.52/mo` net Year-1 w/ `#5e` for new customers) delivered `6.87 GB/s` 1M mixed throughput and `115.87 µs` 4K Q1T1 writes in our Nuremberg lab, while `RS 1000 G12.5` (`€15.22/mo` net w/ `#rs1`) pins 4 physical Zen 5 threads (`0%` steal).
- **Pick Hetzner Cloud (`CAX21` / `CPX31` / `CCX23`)** when you spin infrastructure up and down dynamically via **Terraform, Pulumi, or Kubernetes** and want granular hourly billing in Nuremberg, Falkenstein, or Helsinki.
- **Pick OVHcloud (`VPS-2` / `VPS-3`)** if you host public game servers, VoIP relays, or latency-sensitive multi-continent edges (Gravelines, Frankfurt, Beauharnois Canada) that rely on **OVH's hardware VAC anti-DDoS scrubbing**.
- **Pick Scaleway (`DEV1-L` / `COPARM1`)** if your team operates out of France, Benelux, or Poland (`PAR`/`AMS`/`WAW`) and values consolidated **S3 Object Storage, Managed Databases, and Transactional Email (TEM)** under a single European cloud bill.
- **Pick Hetzner Server Auction (`AX / EX Bare Metal`)** once your footprint grows past 32 GB RAM or requires **Proxmox VE with nested Windows Server / Windows 11 KVM guests** and dual physical NVMe ZFS mirrors.

## Zero-Downtime Contabo Exit & Migration Runbook

Following this three-phase runbook keeps active write downtime under 90 seconds while guaranteeing that stale ISP DNS resolvers lose zero HTTP or TCP requests.

### Phase 1 (T-24h): Lower DNS TTL to `60s` & Run Pre-Cutover Diagnostics

Drop the TTL on all `A`, `AAAA`, and wildcard records to `60` seconds 24 hours before cutover. Then SSH into your Contabo box during evening peak traffic to record baseline `%st`, `w_await`, and true resident memory (`available`, ignoring `buff/cache` — size your target tier in our [Server Sizer](/tools/server-sizer/)):

```bash
# 1. Inspect CPU steal (%st), disk wait (w_await), and active resident RAM
mpstat 1 5
iostat -xz 1 5
free -h

# 2. Benchmark uncached 4K Q1T1 synchronous WAL latency before cutover
fio --name=contabo_wal --filename=/tmp/wal.check --direct=1 --rw=randwrite \
  --bs=4k --iodepth=1 --fsync=1 --size=256M --runtime=15 --time_based
rm -f /tmp/wal.check
```

### Phase 2 (T-1h): Initial Live `rsync` Pre-Sync (Services Stay Online)

Provision your new EU server, install Docker / database engines, and copy 98% of persistent assets over SSH while your Contabo apps continue serving live traffic:

```bash
# Pre-sync compose stacks and static uploads while services remain online
rsync -avzHAX --numeric-ids --info=progress2 \
  /opt/stacks/ root@NEW_EU_IP:/opt/stacks/
rsync -avzHAX --numeric-ids --info=progress2 \
  /var/lib/docker/volumes/ root@NEW_EU_IP:/srv/volumes-staging/
```

### Phase 3 (Cutover Window, ~90s): Stop Writes, Pipe DB Dump, Delta Sync & Proxy Old IP

Pause writes briefly to guarantee transactional consistency, stream compressed dumps directly over SSH, flip DNS, and forward any lingering traffic arriving at the old Contabo IP:

```bash
# 1. Stop application writers on Contabo (keep DB daemon running for clean dump)
cd /opt/stacks && docker compose stop web worker api

# 2A. Stream PostgreSQL custom-format dump directly into the new server
pg_dump -Fc -U postgres appdb | ssh root@NEW_EU_IP \
  "pg_restore -U postgres -d appdb --clean --if-exists --no-owner"

# 2B. (Or for MariaDB / MySQL InnoDB consistent single-transaction stream)
mariadb-dump --single-transaction --quick --routines appdb | \
  ssh root@NEW_EU_IP "mariadb appdb"

# 3. Stop DB on Contabo, run final fast delta rsync, and start stack on new host
docker compose down
rsync -avzHAX --numeric-ids --delete \
  /opt/stacks/ root@NEW_EU_IP:/opt/stacks/
ssh root@NEW_EU_IP "cd /opt/stacks && docker compose up -d"

# 4. Zero-dropped-request shield on old Contabo IP for stale DNS caches (48h)
# Option A (HTTP/HTTPS via Caddy): caddy reverse-proxy --from :443 --to NEW_EU_IP:443
# Option B (Transparent L4 TCP forwarding via iptables):
sysctl -w net.ipv4.ip_forward=1
iptables -t nat -A PREROUTING -p tcp -m multiport --dports 80,443 -j DNAT --to-destination NEW_EU_IP
iptables -t nat -A POSTROUTING -p tcp -d NEW_EU_IP -m multiport --dports 80,443 -j MASQUERADE
```

If you choose Netcup as your target host, eligible non-EU (excl. UK & CH/LI) and EU VAT-ID buyers receive 0% VAT alongside 14-day first-order verification via Stripe Identity or `Vorkasse` prepayment (see our [0% VAT Registration Guide](/guides/vat-free-registration-guide/)), and 12-month voucher plans require 1 month's notice to the end of the month, which you can schedule "at the end of the contract period" in CCP on Day 1 (see our [Contract & Cancellation Guide](/guides/contract-and-cancellation-guide/)).

## Frequently Asked Questions

### How do I verify whether my Contabo VPS is bottlenecked by CPU steal or disk IOPS caps?

SSH into your node during peak European hours (18:00–22:30 CET) and run 'mpstat 1 5' alongside 'iostat -xz 1 5'. Persistent '%st' above 5% proves host CPU contention, while 'w_await' climbing above 15–40 ms during an uncached 4K Q1T1 fio fsync test confirms storage IOPS throttling.

### Which Contabo alternative is best if I want dedicated CPU cores without paying €25+/month?

Netcup RS 1000 G12.5 (€15.22/mo net Year-1 with #rs1; €18.26/mo renewal) pins 4 dedicated AMD EPYC™ 9645 or 9655 (Zen 5) physical threads with 8 GB DDR5 ECC RAM, 128 GB local NVMe RAID, €0 Copy-on-Write snapshots, and 1-click Port 25 unlock — roughly 40% below Hetzner Cloud CCX23 (€24.50/mo net).

### When should I pick OVHcloud, Scaleway, or Hetzner Bare Metal instead of Netcup?

Pick OVHcloud (VPS-2 / VPS-3) when public UDP game servers or Anycast edges require always-on OVH VAC DDoS scrubbing across France, Canada, or APAC; pick Scaleway when you want native Paris/Amsterdam/Warsaw Object Storage and Transactional Email (TEM) in one IAM console; and pick Hetzner's Server Auction (~€34–€42/mo net) when you need physical VT-x/SVM nested virtualization for Windows VMs.

### How do I prevent stale ISP DNS caches from hitting my shut-down Contabo containers after cutover?

After stopping Docker on Contabo and finishing your final database dump + delta rsync to the new server, spin up a lightweight Caddy reverse proxy or kernel iptables DNAT rule on the old Contabo IP pointing to your new EU IP for 24–48 hours until all cached DNS records expire.

---
Source: EUServerLab (https://euserverlab.com/compare/best-contabo-alternatives-no-cpu-steal/) | Open Telemetry Dataset: https://euserverlab.com/benchmarks/vps-500-g12-5.json
