I want to transform a specific instance of a microservice into a standalone offline testing program, similar to common test workloads like those found in stream or specint tests. The purpose is to enrich our own testing library by incorporating personalized workloads alongside the general spec tests.
I am aware of existing trace collection tools like Google's DynamoRIO and Intel's pin-tool, but they mainly generate instruction-level traces and are limited to simulation.
What tools or runtime information should I collect in order to construct an offline workload that closely resembles the behavior of a given on-line program (a specific instance of a microservice)?