01Retried for you
A call that fails is tried again in the same call, on the same route, and the call switches to a backup route where the endpoint has one. You do not write the retry loop.
The state of 237 endpoints on 21 platforms as the service last read it, with the time of that reading: what is live, what is paused, and every incident with what was charged. A call that fails is charged $0.00.
What is live right now on each platform. An endpoint's own page says the same for that endpoint.
30-day availability is the share of time the endpoint was serving calls: at least one of its routes was taking calls, so it was not switched off. It is counted over the last 30 full days in UTC, for each endpoint, and added up per platform and for the whole service. It is not a success rate: how individual calls went is not in it.
Each endpoint's page says whether it is live right now. Browse endpoints →
A call that fails is tried again in the same call, on the same route, and the call switches to a backup route where the endpoint has one. You do not write the retry loop.
A route that keeps failing still takes calls: a request the source cannot serve for one customer never cuts off anyone else. A failed call is not charged, and goes on to a backup route where there is one. An endpoint answers 503 capability_paused only when we switch it off by hand, and this page says so.
A failed, timed-out or uncertain call costs nothing: what it held is released at once. Every incident below says what was charged.
What broke, for how long, and who paid for it.
Stable parts only grow. What changed for callers is written here.
Nothing to change; the addresses you read are sharper.