Rally uses timers internally to measure when a request started and when it has completed. However, it has stopped measuring as soon as the first byte of the response has been received instead of waiting for the full response. This can skew service time measurement if the response is large and Elasticsearch streams it.
With this commit we wait until the last chunk of the response has been received. This results in a more realistic service time metric in these cases.
Rally uses timers internally to measure when a request started and when it has completed. However, it has stopped measuring as soon as the first byte of the response has been received instead of waiting for the full response. This can skew service time measurement if the response is large and Elasticsearch streams it.
With this commit we wait until the last chunk of the response has been received. This results in a more realistic service time metric in these cases.
Closes #1822