The http_endpoint input currently has no mechanism to be able to communicate back pressure to a client. This has resulted in OoM failures in user systems when the input's outflow cannot be sufficiently rapidly ingested, due to queue memory consumption.

We can communicate back to the client that we are unable to continue to receive events by using a 429 HTTP status, but we need to be able to asses the amount of in-flight data from the client to the index. We can achieve this by watching the input metrics to keep the number of received events minus the number of ACKed events less that some value, also considering the expected size of messages from the client. It may also be possible to achieve this by keeping a running sum of in-flight message length.

Classifying this as a bug rather than a feature request since it is causing loss of service for users.