What's the easiest way to "profile" servers in production?

Question

PostgreSQL

How to profile a web service with minimum effort?
I tried to deal with cProfile, but something is not right.
What would I like? To see specifically in a pleasing way (on charts with tables, for example), how long it took to process the request as a whole, how much time was spent in the python code, how many requests there were in pg, how long they took, how long requests for radishes took, etc.

What doesn't exist?

Answer 1

newrelic.com
else
www.fullstackpython.com/monitoring.html

Answer 2

Easiest - no way. All the same, profilers steal well in the performance themselves. Or information is scarce. Logs are quite enough here, but carefully test suspicious places on a dedicated server. Yes, even on a virtual machine locally, or even locally.
It exists, yes. Only you wrote very little. Google it like this . But still it is better to turn it off, turning it on only at moments of relative downtime. Which is also not very true, but inadvertently screw up the car due to completely unnecessary IO operations.
UPD.
cProfile is python. Unfortunately, despite the rather rich experience of python development, I have never profiled it. It's just not necessary - it's too slow and all the really bottlenecks are visible to the naked eye, so to speak.
UPD2.
A friend just unsubscribed, says "try z3c, but it's zope." Another, flask'er: "still looking, but for now debugtoolbar, almost like on django". In general, a lot depends on the chosen framework.
UPD3.
The system administrator tells Zabbix , though it will be necessary to shaman a little with your hands in order to make and / or find all the widgets, but the result is very good. steep.