Theoretically this should fix the SocketError: other side closed errors we are seeing.
Based on the theory that the body of a request must be consumed or destroyed it seems to me that for requests which yield a HTTP 200 (successful) response we are consuming the body while for request which yields a non HTTP 200 (unsuccessful) response we are not.
Non HTTP 200 responses do have a body too.
This consumes the body on non HTTP 200 responses. Consuming the body over destroying the stream is done because destroying the stream will trigger a socket error which causes a termination of the request differently that what we want when the upstream service is yielding http errors.
Theoretically this should fix the
SocketError: other side closed
errors we are seeing.Based on the theory that the body of a request must be consumed or destroyed it seems to me that for requests which yield a HTTP 200 (successful) response we are consuming the body while for request which yields a non HTTP 200 (unsuccessful) response we are not.
Non HTTP 200 responses do have a body too.
This consumes the body on non HTTP 200 responses. Consuming the body over destroying the stream is done because destroying the stream will trigger a socket error which causes a termination of the request differently that what we want when the upstream service is yielding http errors.