Provider clusterDay-zero acceptanceUpdated 2026-11-06

Kamatera setup: configure CPU type, verify the order summary, then harden SSH

Select deliberately in the live calculator, reconcile the order summary, and accept the host before valuable data arrives.

Workload before headline price — verify the plan, terms and recovery path
How to use this. Steps reflect common operator problem-handling patterns and public Kamatera flows. Follow your live console labels; discover actual OS, services, and paths on your host before production changes.

Fresh / occupied gate and prerequisites

Use this guide for a new server. If the host already holds data you care about, stop and complete control writes → final export → verified restore before rebuild or terminate.

Private prerequisites: monthly vs hourly, Type A/B/D/T hypothesis, vCPU/RAM/disk, DC preference, traffic package expectation (especially Asia 1 TB), OS vs marketplace app, backup/managed checkboxes, budget, and a terminate calendar for trials. The reader performs purchase in the Kamatera console.

Build the server in the live calculator

Order path checklist (follow live labels):

FieldVerifyFail if
DatacenterUsers/data residency; confirm code appears liveAssuming marketing 25 equals stock
CPU typeA / B / D / T semantics from the reviewType T without overage monitoring
vCPU / RAM / NVMeHypothesis from scenario tableTreating starter cards as the only SKUs
Traffic package5 TB vs Asia 1 TB vs unmetered 50 MbpsInventing overage math
OS or marketplace appUbuntu vs n8n/WP/Ollama/OpenClaw imageAssuming hardened production defaults
Extended daily backup / ManagedOptional lines; managed ~$50/mo seed in captured calculatorEquating managed with ScalaHosting WP Manager
Monthly vs hourlyCash mode intentionalLeaving hourly leftovers without terminate date

Type T reminder: calculator UI is hourly-only; burst overage risk is real.

Read the order summary like a bill

  • Confirm traffic package matches the DC story you intended.
  • Confirm Extended daily backup and Managed lines if checked.
  • Confirm billing mode; if Type A, keep the FAQ vs calculator hourly conflict labeled—do not “fix” the rate.
  • Save a private screenshot/PDF of the summary; never publish secrets.

First login: discover, then harden

Marketplace default credentials and first-login panels are UNKNOWN without deploy—discover what the console actually delivered.

  1. Read-only inventory: OS version, listening ports, containers/services, mounts, public IP, any cloud firewall state.
  2. Create a sudo admin with SSH keys; prove a second session works; then narrow password/root login per your policy; keep console recovery.
  3. Host firewall: intentional 22/80/443 unless an app requires more—document exceptions.
  4. If a marketplace image is present: record admin URL, ports, and version paths; do not leave admin UIs world-open.
# Placeholders only — discover real users/paths on your host.
ss -tulpn
df -h
docker ps --format 'table {{{{.Names}}}}\t{{{{.Ports}}}}' 2>/dev/null || true

Day-zero acceptance receipt

Acceptance checklist: billing mode recorded; CPU type/size; DC; traffic package; add-ons; SSH key-only working; firewall map; image/app inventory; optional snapshot/image before first risky change; off-host export destination named.

Handoff: n8n operations, AI apps, billing / backup / cancel.

Stop conditions: unexpected public listeners, reused root passwords, unintended paid add-ons, Asia traffic mismatch on the summary.