Middleware auto instrumentation #3483
Labels
community
Was opened by a community member
feature-request
A request for a new feature or change to an existing one
Hello everyone 馃憢
Middleware over the long term can have a significant impact on latency. I think it would be a good addition to materialize the time spent on it. Especially, as it's not inferable by making a difference between
rack.request
andrails.action_controller
because of #2100.I'm currently trying to implement this in our codebase, and so far I've come up with 3 ideas:
For 1. the implementation is simple, but this create a huge pyramid. Flamegraphs become complicated to navigate. And you need to do the difference yourself to infer real execution time. But in opt-in mode it's a start to aggregate some data on a small amount of time.
For 2 and 3, Implementation is a little more complicated, but it's the best I've come up with so far.
I'd love to hear your feedback, suggestions and ideas about all this. 馃檹
The text was updated successfully, but these errors were encountered: