zalando-zmon / opentracing-utils Goto Github PK

Add examples for how opentracing-utils can be used, for instance with multiple libraries.

The libraries of the specific vendors (lightstep, instant, jaeger) are not part of dependencies.txt of opentracing-utils and have to be added to the dependencies of the application which is using opentracing-utils (otherwise it will fail, e.g. here https://app.eu.scalyr.com/y/7pqX2nx43k2). It would be nice to mention it in the README in the Usage part.

Hello,
First of all, thank you for maintaining this project. We are using it and it is very helpful.
Second, I would like to know if it is possible, and is it a good practice, to log the parameters of a SQL query along the query itself (in the trace_sqlalchemy() function in _sqlalchemy.py . It might be very useful for debugging purposes.

The main downside I could think of is the possibility of logging private information of user's input.

I would like to know what do you think about it.

Thanks,
Avi

opentracing==2.0 introduced scope manager API. The tracer is responsible for scope_manager implementation.

The order for detecting parent spans needs to be updated. I suggest one of the following two orders:

Scope manager first

1. Using opentracing.tracer.active_span managed by the tracer context manager. The new span will be using the
   scope manager.
2. Using span_extractor.
3. Detecting span in kwargs.
4. Using call stack frames inspection.

Explicit first

1. Using span_extractor.
2. Detecting span in kwargs.
3. Using opentracing.tracer.active_span managed by the tracer context manager. The new span will be using the
   scope manager.
4. Using call stack frames inspection.

as span_extractor and passing spans in kwargs is more explicit.

Can be added to OpenTracingHttpMiddleware and using Database instrumentation support in Django>=2.0

Lots of bits are missing

Example:
Tests that combines Flaks, requests and SqlAlchemy tracing.

Python2 EOL is coming soon, and we should support Python3-only and move on!

Does this library support spans across multiple libraries?
Like using flask and sqlalchemy within the same trace?

def test_nested_libs(self):
    engine = create_engine('sqlite://')
    User.metadata.create_all(engine)
    session = sessionmaker(bind=engine)()
    app = Flask('test_app')

    @app.route('/')
    def root():
        user = User(name='Tracer', is_active=True)
        session.add(user)
        session.commit()
        return user

    trace_flask(app)
    trace_sqlalchemy()

    with app.app_context():
        print app.test_client().get('/').data

With 1 flask trace with a child sql span

In certain cases an opentracing-utils user might need to skip @trace spans based on custom logic.

Suggested usage example:

def skip_span_logic(*args, **kwargs):
    if  'something' in kwargs:
        # @trace should not add a new span
        return True

    # @trace should add a new span
    return False


@trace(skip_span=skip_span_logic)
def trace_me(*args, **kwargs):
    # do stuff

This could be helpful in the case of trace_requests(ignore_url_patterns=[...]), since the skip_span func could inform @trace whether to add a span or not.