[Question] spanning across sub-processes

[thehesiod]

In our application we have a process pool of botocore workers. Is there a way, in general, to ensure that sub-processes can chain up to their parent process traces?

I'm guessing the right context details needs to be transferred to the sub-process to make things work.

[thehesiod]

btw for asyncio I found I could do it by serializing the parent span in the base process via:

        dd_span = tracer.get_call_context().get_current_span()
        attrs = {'name', 'service', 'resource', 'span_type', 'trace_id', 'span_id', 'parent_id', 'start'}
        dd_span_info = {n: getattr(dd_span, n) for n in attrs} if dd_span else None

then smashing it in in the child process:

try:
        if dd_span_info:
            dd_ctx = tracer.get_call_context()
            parent_span = Span(tracer, **dd_span_info, context=dd_ctx)

            dd_ctx._current_span = parent_span

            task = asyncio.Task.current_task(loop=self._loop)
            set_call_context(task, dd_ctx)
finally:
            if dd_span_info:
                set_call_context(task, None)

This works because the tracer only needs a few key pieces of information. If we could formalize a process for this it would be awesome!

[thehesiod]

hmm, reading http://pypi.datadoghq.com/trace/docs/#distributed-tracing it seems like perhaps the asyncio patchers should support parent trace/span IDs

[thehesiod]

ok, I think I've implemented something acceptable in my PR: #248 It adds optional kwargs that follows the aiohttp style for distributed spans...however if there was a way to officially tell the context what the parent was it would be better I think.

[palazzem]

Related to #247 that has been solved. Closing this one too.