Check implication of __init__ for rabbitmq span type

[instana]

Check implication of __init__ for rabbitmq span type

[andrewslotin]

@manojpandey, could you verify if the exit RabbitMQ spans have proper kind?

[pglombardo]

Because rabbitmq is a dual type span (entry & exit) based on the sort key it's registered in both ENTRY_SPANS and EXIT_SPANS.

So the default exit kind for all spans is 1 (entry):

python-sensor/instana/span.py

Line 245 in f800501

self.k = 1

Then through the if statement the exit type condition gets hit (because rabbitmq is a member of ENTRY_SPANS):

python-sensor/instana/span.py

Lines 250 to 252 in f800501

	elif span.operation_name in self.EXIT_SPANS:
	self.k = 2 # exit
	self._populate_exit_span_data(span)

Then we set it back to exit type if the sort key matches:

python-sensor/instana/span.py

Lines 257 to 258 in f800501

	if "rabbitmq" in self.data and self.data["rabbitmq"]["sort"] == "consume":
	self.k = 1 # entry

The path of the logic is solid although it's not that evident and could possibly be reorganised to be easier to understand. rabbitmq is special case because of the kind duality.

It looks like the tests are not validating the kind type so ideally that should be added to the test runs:
https://github.com/instana/python-sensor/blob/master/tests/clients/test_asynqp.py

[andrewslotin]

Hey @pglombardo, great to see you again! I see the point, however the problem is that rabbitmq is listed both in ENTRY_SPANS and EXIT_SPANS, and while processing it we'll hit the branch that uses _populate_entry_span_data() and keeps the self.k unchanged, so by the time the if "rabbitmq" in self.data and self.data["rabbitmq"]["sort"] == "consume" is executed, self.k will be still set to 1.

[pglombardo]

Vacation has obviously made me weak. I see the bug now. :-)