I wrote something out of our need for better monitoring/metrics for Nginx here: https://luameter.com/
And quite a few people are using it for per-vhost metrics use case just like you said. Also if you prefer something simpler and open source I also wrote this https://github.com/lebinh/ngxtop
I showed ngxtop [0] about 2 years ago as a top inspired tool to quickly check the status of my nginx servers, something I created for my need. I expected some people would find it useful and shared, turned out a lot of people need something to monitor their nginx servers and it got 3k stars in a day or two. So I built a much better tool to help monitor nginx and sell it as a product, Luameter [1]. It didn't take off like the other but now making a couple hundreds a month and being used by some high-profile users, i.e. well-known companies.
About a year ago I open sourced a tool for quick monitoring and troubleshooting Nginx, ngxtop [1], and received some very positive interests from HN community. However quite a few people are using it as a monitoring solution for their Nginx server, which is not ideal. Ngxtop was designed to be like the `top` command, you fire it up to check for at most 5-10 minutes and stop. It was not meant to be a long monitoring process that can be used to plug to your monitoring system such as Zabbix or Nagios or graphed by Graphite. Nginx has an official status module for this [2], but it requires Nginx Plus, a > $1k product. Other solutions would normally require a log processing process piping to an external system to calculate and store the metrics. While an ELK stack [3] could provide you a lot more useful data, its setup and maintenance efforts can be significant. And that's why I created Luameter.
It easier to setup, flexible to configure, and equally easy to be plugged into external system given its JSON api.
Unfortunately most of the servers I'm working with are still on python2 by default. The latest Ubuntu LTS, 12.04, only has python2 by default so python2 is still my priority. But in this case, a port should be trivia so I'll look into python3 as soon as possible.
Yes, ncurses would be much better but I haven't had time for it yet. Clearing screen is simple enough and work pretty well for me so far so here we are :)
If what you mean by exec time is request_time, i.e., serving time of a request, then it's pretty close on my agenda as long as you have it in your access log :)
The performance impact really depend on your server and traffics. Currently with high traffic server it will take quite some amount of CPU if running for awhile, after 15-20 mins or so, but I'm working on reduce that. So now it mostly useful to know what's going on with your nginx at the moment but not over a long period.