Previous incidents

The ability to perform javascript-client evaluations and fetch configurations were down.
Many configuration loads would have been satisfied via our CDN during this time so only some customers were affected.

We've investigated and fixed a few issues to prevent this in the future
* improved anti-affinity rules to spread API service instances into more availability zones
* installed a missing set of Load balancer health checks for the API

50 seconds of downtime for the evaluation endpoint. New visitors to JS clients received fallback values. Existing visitors got cached values.

30 seconds of downtime for the evaluation endpoint. New visitors to JS clients received fallback values. Existing visitors got cached values.

3:18 - We're investigating API stability issues.

3:43 - We've resolved the issues.

We had issues scheduling new pods due to an overly restrictive anti-affinity policy. This exacerbated a deployment issue.