Hi,
I am trying to measure response latency of some lua code running in production, and so I tried to analyze ngx.var.request_time values.
However, ngx.var.request_time seems to be calculated as total request handling time, which includes a time of I/O operations like response delivery to the client as well as LUA code execution time.
In my case, some of the responses could be quite slow due to low network performance of the clients (mobiles), and so my statistic figures are affected by actual client behavior very much.
So my question - is there any good way to get a request handling time, excluding a time spent on client related I/O ?
Thanks
BR,
Pavlo