Run source code deterministic profiling tests

The Profiler plugin can be used whenever you make changes to grafana-bridge to get a quick sanity check on the overall performance of the source code. It can help you answer questions such as how many times a particular function was called, or how much total time was spent within that function. The simplest way to create a profile runtime statistics dump of a particular function is to call the Profiler.run() method in the corresponding function unit test.

profiler = Profiler()
opentsdb = OpenTsdbApi(logger, md_instance, '9999')
resp = profiler.run(opentsdb.format_response,*(data, jreq))
assert resp is not None
assert os.path.exists(os.path.join(profiler.path, "profiling_format_response.prof"))

It is also possible to create a runtime statistics dump in the production environment during operation. This can be achieved by setting the runtime_profiling variable to True in the analytics.py file during the startup of the grafana-bridge application. For all methods that have been decorated with the get_runtime_statistics function, the Python deterministic profiling set will be called.

@get_runtime_statistics(enabled=analytics.runtime_profiling)
def __parseHeader(self):
    item = self.json['header']
    return HeaderData(**item)

A profile dump for each method will be written to a file with the signature 'profiling_.prof'. Each method execution results in the overwriting of the file with a new dump.

The statistics report over all captured dumps can be generated and accessed via the /profiling endpoint.

However, this should only be done under supervision and for a short period of time. Furthermore, the Profiler plugin adds a noticeable runtime overhead due to the additional instrumentation code that needs to register and track certain events. Especially in a production environment that is already suffering from poor performance, the constant collection of profiler statistics can put the application into a critical state.