Service level

Availability target

We target 99.9% monthly availability of the game port for each instance, measured at five-minute intervals from two external checkers in different networks. Voice relays have the same target on their advertised port.

99.9% allows about 43 minutes of unavailability per month. We publish the figure because it is honest for a single node with a shared uplink; a promise of 99.99% from a provider of our size would be marketing rather than engineering.

What counts as unavailability

  • Two consecutive failed checks on the game port from both external checkers.
  • Unplanned maintenance performed without notice.
  • Loss of access to the panel for more than fifteen minutes.

What does not count

  • Planned maintenance announced at least 48 hours in advance.
  • Failures caused by a game update, a broken mod, or a configuration change made by the account holder.
  • Outages of the game's own services, including Steam, PlayFab and official authentication.
  • A DDoS attack large enough to require blackholing in the region, as described on the network page.
  • Unavailability of a voice channel caused by client-side network conditions.

Maintenance

Maintenance windows are 02:00–05:00 local node time. Most work — storage refreshes, kernel updates, panel releases — is done without stopping game instances. When a restart is unavoidable, we announce it at least 48 hours ahead by email and in the panel, and we schedule it outside the peak hours of that region.

Credits

Monthly availabilityCredit
99.0% – 99.9%10% of the monthly fee
95.0% – 99.0%25% of the monthly fee
Below 95.0%50% of the monthly fee

Credits are applied to the next invoice, are not paid out in cash, and are not available on accounts with unpaid invoices. To claim a credit, write to sales@amber.cfngg.net within thirty days of the incident, quoting the affected instance. We do not require you to prove the outage; we have the checks.

Support response

Median first reply during 09:00–21:00 EET is under 30 minutes. Outside those hours the on-call engineer responds to incidents affecting multiple customers; single-instance requests are answered the next working morning. A written estimate is given for any repair that will take longer than an hour.

Backups

Backups are a component of the service, not a guarantee against data loss. We verify a random restore every week and keep the retention stated on the pricing page. If a backup we hold is unusable, the credit table above applies to the affected month, and we help reconstruct the world from the most recent usable snapshot.

Last updated February 2, 2026.