99.5% per instance,
measured, not promised

Service credits

Credits are a percentage of the compute charge for the affected instance in the month the availability fell short. They are applied to your next invoice and never exceed that instance's charge for the month.

To claim, open a support ticket within 30 days of the end of the month, naming the instance. We reconcile against our own monitoring — you do not need to prove downtime we already recorded.

How availability is measured

• Our monitoring probes every running instance from inside the region every minute. A minute in which the instance does not answer counts as unavailable.
• Availability for the month = (total minutes − unavailable minutes) ÷ total minutes, per instance.
• Minutes during which the instance was stopped by you, or by us at your request, do not count either way.
• The same probes feed the public status page, so the number you see there is the number credits are calculated from.

What is excluded

• Scheduled maintenance — announced at least 72 hours ahead on the status page, and held in the Sunday 22:00–24:00 IST window unless a security fix cannot wait.
• Things inside your instance — the operating system, your application, a full disk, a firewall rule you set. We restart a stopped instance automatically and tell you; we do not manage what runs inside it.
• Your upstream — DNS you control, a domain that expired, a payment that lapsed and suspended the account after the notices described in the terms.
• Force majeure — events outside the datacentre operator's reasonable control.

What we tell you upfront

Each region runs on dedicated bare-metal today, one host per region. A host failure means a coordinated restore from snapshot, typically under 30 minutes, and it counts against the SLA. A multi-host tier with a 99.9% commitment is on the roadmap; we will publish it here when it exists, not before.