Deployments and uptime

What an API client sees during a deploy or database maintenance, how an unavoidable interruption is announced, and the availability in the Terms of Service.

Beliq deploys new versions and maintains its servers while the API is in use. This page says what a client of the API sees while that happens, and what it should do.

A deployment replaces the servers that answer the API one after another. A new server takes requests before an old one stops, and an old server finishes the requests it is already working on before it goes.

So a deployment has no maintenance window and is not announced. Your integration needs no special handling for one, and there is nothing to pause.

Routine work on the database, such as a restart for an update, hands its work over to a second database server. While that happens, a request that needs the database is held until the second server has taken over, and is then answered.

None of this is a promise that no request ever fails. Treat a failure during a deployment or during maintenance like a failure at any other time. The error reference says which errors are worth sending again and how long to wait first. It also says when the quota unit of a failed request goes back.

Under section 4.2 of the Terms of Service, we carry out maintenance and updates in a way that keeps the service reachable while they run. Where an interruption is exceptionally unavoidable, we announce it ahead of time, place it overnight and keep it as short as possible. The section has the notice period and the hours.

One kind of interruption is not announced: a measure needed to avert an acute risk to security, data integrity or operations can be carried out at any time, and we tell you immediately afterwards.

A server, a data centre or a network link can fail with no warning, and requests can fail until the servers that are left have taken over. The status page shows the current state of the service and its incident history. An independent monitoring service runs it and checks Beliq from outside our network. It checks every few minutes and counts an outage only after several failed checks in a row, so an interruption shorter than that does not show in the availability figure on the status page or on beliq.eu.

Section 4.1 of the Terms of Service says we owe an API availability of 99.5% on a monthly average. The section also says where that is measured and what counts as downtime.

For the free plan and the tools on beliq.eu that need no account, the Terms for Free Offerings take precedence over section 4 of the Terms of Service, and they give no availability commitment.