Continuing the trace

[kamalhussain]

I have a scenario where the end-to-end flow is broken since there is a queue involved. As an example, the client makes a request to a web service, which stores the request in a queue. Another process reads from the queue and makes a new request to complete the flow. I would like to trace the complete flow and wondering how to achieve it. Here is the general approach I've in mind.

1. At the end of the first hop, I capture the current context and stores in the queue.
2. At the beginning of the second hop, the context is read from the queue and used for restarting the trace.

I run into a couple of issues while trying to implement it. For example, capturing the context and saving gives me the following error:

# read the context
current_context = context.get_current()
# serialize data
serialized_data = pickle.dump(current_context)
# store serialized_data in the queue
..

Here is the error:

TypeError: can't pickle _thread.lock objects

If there is a way we can reliably serialize the context, I am thinking of using the following approach to continue the trace. However, I'm not able to test since I am stuck on the first error.

    # read the context_data from the queue
    ...
    # de-serialize the context_data
    saved_context =  pickle.loads(context_data)

    with tracer.start_as_current_span("redis", saved_context)
        ... 

Any pointers would be appreciated.

Thanks

[kamalhussain]

I could solve this issue by using inject() and extract() methods instead of using pickle module.