We’ve added a near-real-time status monitor that tells you whether each prediction service in the network is online and producing proper results.
It works by sending a small query to each prediction service once every ten minutes, and making sure that each service returns a result within five minutes that complies with its WSDL. Therefore, in order to show a green light, a service must be (a) serving a valid WSDL and (b) producing results that conform to the schema within that WSDL. The page then shows the server status at the time the last test query was sent.
Any known problems with the predictors will be shown on that page too, as and when we discover them.
Coming soon: We’re going to add functionality to that page to check the front-end service — i.e. the pipeline as a whole — too.
