Changelog

Encrypted disaster-recovery backups

Backups are now full, self-contained copies of a disk shipped off the cluster to independent object storage, so they survive losing the cluster, not just a bad change. Every backup is disaster recovery by design: you pick the destination backend (Wasabi in Germany, ImpossibleCloud in France, or local ServersCamp S3), the archive is compressed with zstd and encrypted before it ever leaves the host with your organization's key (age, ChaCha20-Poly1305), and you can view that key in the cabinet to decrypt archives yourself. Two SHA-256 fingerprints are recorded per backup (the raw disk and the stored archive); the raw hash is re-verified on restore before anything is written, so a corrupt restore is refused rather than booted.

Snapshots vs backups, explained →

Automatic backups with GFS retention

Schedule automatic backups per organization (and per VM or disk) from Backups → Auto-backup. Off by default; turn it on with a retention preset: Basic (keep 3 daily), Weekly (7 daily + 4 weekly), or full GFS (7 daily + 4 weekly + 12 monthly + 3 yearly), plus a time window. Auto-backups go to a DR backend or to Random, which spreads each run across Wasabi and ImpossibleCloud. The scheduler is concurrency-capped and runs in small batches, so a fleet-wide backup window never hammers your VMs.

Backups: redesigned detail page

Each backup now has a structured detail page: general info, source VM or disk, storage and savings, recovery objectives (RPO measured as the real age of the copy, RTO as a clearly labelled estimate), encryption and format, and an integrity panel with both SHA-256 checksums you can copy. The backup list also gained filters and sorting inside the table, plus a destination chip showing which backend and country each archive lives in.

Virtual Machines: reinstall from a new OS

The new Reinstall tab on a VM lets you wipe the root disk and reinstall a fresh OS (same template or a different one - Ubuntu, Debian, Rocky, Alma, CentOS Stream, Alpine, and more). VM identity, name, networks, IPs, MAC, security groups, attached volumes, floating IPs and login credentials all stay put; only the root filesystem is replaced. Pick the template, type the VM name to confirm, choose between graceful or force stop, and watch the progress live (stopping → wiping → cloning → booting). Existing snapshots and backups of the reinstalled VM become orphaned, as expected.

Virtual Machines: resize is live

You can now change a VM's compute class at any time from the new Resize tab on the VM page. Pick any active class in Burst or Dedicated mode, see a live diff of vCPU, RAM, mode and price, then apply. Mode toggle (Burst ↔ Dedicated, same vCPU and memory) is applied live with zero downtime. Size change (different vCPU or memory) requires a short restart - you pick between graceful ACPI shutdown or instant force stop in the same screen. Progress is shown step by step right in the tab; if the page is reloaded mid-resize, it resumes the live view automatically.

Read about compute classes →

Virtual Machines: refreshed creation wizard

The VM provisioning wizard got a redesign. Tighter step layout, clearer compute class picker with explicit Burst/Dedicated explanation and a link straight to the new Compute classes docs page. Storage classes now fit in two rows on a single screen with bandwidth and IOPS bars; network classes share the same visual language. Step 5 (Review) is rebuilt as a proper summary with detail cards and a cost sidebar that mirrors the footer total exactly. If you leave the name field blank, the placeholder is now used as the actual VM name. The hardware class "legacy" is presented as "Basic" in the picker.

Virtual Machines: redesigned list page

The VM list page in the console got a full refresh. New header summary with four KPI cards: fleet load (avg CPU and disk utilization plus the peak VM), allocated capacity (vCPU/RAM/disk totals), recent activity feed, and public network usage for the current month with live bps. The instance table is now full-width with denser rows that surface image, class tier, allocation, public/private IPs, runtime state, per-VM load bars, SSH key and CPU burst setting at a glance. Under the hood, the list endpoint now denormalizes class and key names so the page loads in a single round-trip instead of fanning out across five separate API calls.

Container Registry: first release of our managed registry

Managed Container Registry is live. Private OCI registry on a dedicated VM (one registry = one tenant = one scrd instance), built on top of CNCF Distribution v3 with our own daemon (scrd) wrapping auth, RBAC, Trivy vulnerability scanning, lifecycle policies, 90-day audit log, webhooks and pull-through cache. Ten fixed tiers from €3.02/mo (10 GB disk, 200 GB included egress) up to €85.03/mo (1 TB disk, 32 TB included egress). Every capability enabled on every tier.

Read the product page →

MCP server: connect Claude to your pricing data

Public read-only MCP catalog at https://platform.serverscamp.com/mcp. Six tools cover VM, storage, network, managed PostgreSQL, bucket classes, and extra prices (Public IP, Floating IP, snapshots). Add it as a custom connector in claude.ai, Claude Desktop, Cursor, or any MCP-compatible client and ask pricing questions in plain English: "cheapest VM with 8 GB RAM", "compare PostgreSQL classes under €50/mo". Account-scoped tools (list your VMs, create one, attach a volume) are next.

Read the docs →

First public benchmark: AWS RDS for PostgreSQL vs ServersCamp

We published a side-by-side pgbench TPC-B benchmark of managed PostgreSQL on AWS RDS db.m6i.large vs ServersCamp db-m-8g. Same hardware, three database sizes (in-cache, partial cache, disk-bound), three runs each, on-demand pricing. Headline at storage parity: ServersCamp leads disk-bound by +9% TPS, cuts p95 latency in half, runs ~5× more stable on tails - at ~½ the price. Full methodology, raw pgbench output, and honest CPU-generation disclaimer included.

Read the writeup →

Open preview: platform live, free while we wire up payments

The platform is up and most services are running: VMs, volumes, networks, floating IPs, security groups, S3 buckets, managed PostgreSQL, container registry. The billing engine is running too. You can see live cost meters per resource in the console. What is not live yet: payment processing. Until that lands, sign-ups stay open and usage stays free, no credit card needed, no surprise invoice. We'll announce the cut-over with plenty of notice.

Managed PostgreSQL & CPU guard

Two major launches today:

• Managed PostgreSQL - first release of managed PostgreSQL service with automated backups and point-in-time recovery.
• CPU Guard - fair CPU usage enforcement for burst-type VM classes. Burst instances share CPU resources and now have transparent usage limits to ensure consistent performance for all tenants.

VM backups

Implemented automated backups for virtual machines. Create snapshots manually or schedule automatic daily/weekly backups with configurable retention policies.

Web console for VMs

Access your virtual machines directly from the browser via VNC protocol.

Control panel updates

New features in the control panel:

• Support tickets page: submit requests directly from your dashboard
• VM details page: new tabs for attached disks and networks

VM management improvements

Major updates to the control panel:

• VM power controls: reboot, hard reboot, shutdown, hard shutdown
• Rename VM display name on the fly
• New detailed VM information page
• Tag management for VMs: view, add, and remove tags

New Linux templates

Added latest Linux distribution templates:

• Debian 13
• Rocky Linux 10
• Alpine 3.21

Terraform provider v1.2.0

Released official Terraform provider for ServersCamp. Manage your infrastructure as code with full support for:

• SSH Keys
• Virtual Machines
• Routers
• Networks

View on Terraform Registry →

CLI v0.9.2: VM creation timing output

Added detailed timing breakdown when creating VMs with --debug flag. Now you can see exactly how fast your VM is provisioned:

• Control request accepted: ~200ms
• VM provisioning: ~1.3s
• VM running: ~3.1s
• VM reachable (ping): ~8.5s

Quick start guide and API documentation

Published comprehensive Quick Start guide with step-by-step instructions for deploying your first VM. API Reference now includes code examples in 8 languages: cURL, Python, Node.js, PHP, Go, Ruby, Java, and Rust.

Premium storage class: 25,000 IOPS

New premium-storage tier for I/O intensive workloads. Sustained 25,000 read/write IOPS with 1,000 MB/s throughput. Perfect for databases, analytics, and high-frequency trading.

IPv6 support for all VMs

All virtual machines now receive both IPv4 and IPv6 addresses by default. Public IPv6 addresses are included at no additional cost.

CLI v0.9.0: S3 commands

New scli s3 commands for object storage operations: ls, put, get, rm, cat, mkdir. Works with any bucket, credentials auto-configured.

Live VM class switching

Upgrade or downgrade VM class without stopping your instance. CPU and RAM changes apply within seconds with minimal impact.

Object storage (S3-compatible)

Launched S3-compatible object storage. Create buckets, upload files, generate presigned URLs. Full compatibility with AWS SDK, s3cmd, and rclone.

Public beta launch

ServersCamp is now open for public beta. Free tier available: 1 vCPU, 512 MB RAM, 10 GB NVMe for 6 months. No credit card required.

CLI v0.8.0: initial release

Released scli command-line tool. Manage VMs, SSH keys, and storage buckets from your terminal. Available for macOS, Linux, and Windows.