One of the things we have noticed as we increase our test coverage, is that in addition to the poor quality of the documentation, there a lot of randomness in the errors we get sent back. Today’s case in point being the Yahoo BOSS API.
Configuring the test cases required enabling billing support which we were fine with, and for the levels of use we’re going to have, it won’t be a huge amount. Once set up, the first query went through just fine, returning a HTTP 200 and a nicely formatted block of JSON with the search data enclosed.
Having satisfied ourselves that the test was working as expected we scheduled the test to run every hour.
We then started getting 503 errors from the test, with the specific error code “Key has exceeded its configured rate limit.”
Firstly, the documentation for the API is clear that as a paid for API, this API does not have pre-set rate limits. This makes this hard to debug.
Secondly, and more, in our mind importantly for Yahoo, why is that error coming back as a 503 (Service Unavailable) error and not a 4XX error – we’d typically expect rate limit errors to give back a 403 error or similar.
One of the key services we see APIContext as giving to API providers isn’t just regular testing of the end-to-end experience but also showing up when incorrect or poor errors are being reported.
Sign up for an account today!