Enrich Tracing in Python

Manual Instrumentation with Open Tracing

Configuring Thundra’s Trace Support in the Python agent involves declaring a trace object at the start of your Lambda function. Trace Support is enabled by default when you wrap your function; however, if you would like to configure Trace Support and also perform manual instrumentation, then you must get a tracer instance with the command ThundraTracer.get_instance()

Tracing Code Blocks Using Context Manager

To trace a code block, you can use the context manager returned calling

tracer.start_active_span(operation_name="write_user_to_db")

where operation_name indicates the name of the span relevant to the traced block of code.

from thundra.opentracing.tracer import ThundraTracer
​
def save_user(name, surname, id):
    # Get tracer instance
    tracer = ThundraTracer.get_instance()
    # Trace code block with context manager
    with tracer.start_active_span(operation_name="write_user_to_db") as scope:
        # Perform operations to be traced
        # ...
        

If you already use OpenTracing to instrument your code, you can overwrite the opentracing.tracer with a Thundra tracer instance:

# Get tracer instance
tracer = ThundraTracer.get_instance()
# Set the tracer
opentracing.tracer = tracer

An example usage is show below:

from thundra.opentracing.tracer import ThundraTracer
​
def save_user(name, surname, id):
    # Get tracer instance
    tracer = ThundraTracer.get_instance()
    opentracing.tracer = tracer
    
    span = tracer.start_span(operation_name="operation_name")
    span.set_tag('key', 'value')
    # Do some operations ...
    # ...
    span.finish()

Tracing Methods Using @Traceable Decorator

To trace a method using automated instrumentation, import Traceable from Thundra’s trace plugin, as shown below:

from thundra.plugins.trace.traceable import Traceable

Then, you can use the @Traceable decorator to trace your functions:

@Traceable(trace_args=True, trace_return_value=True, trace_error=True, trace_err)
def method_to_be_traced():
  """ Traced method """
  # ...

Offline Debugging

Offline debugging tracing is supported by Thundra's Python agent, and it allows you to monitor your Lambda functions by each line of code. By default, offline debugging is disabled, but can be enabled by passing arguments to the @Traceable decorator.

Offline debugging can be enabled by setting trace_line_by_line to True.

@Traceable(trace_line_by_line=True)
def method_to_be_traced():
  """ Traced method """
  # ...

Local variables in the method can be traced by setting trace_local_variables to True when offline debugging is enabled.

@Traceable(trace_line_by_line=True, trace_local_variables=True)
def method_to_be_traced():
  """ Traced method """
  # ...

Source code in lines can be traced by setting trace_lines_with_source to True when offline debugging is enabled.

@Traceable(trace_line_by_line=True, trace_lines_with_source=True)
def method_to_be_traced():
  """ Traced method """
  # ...

Configuring Trace Arguments

By configuring trace arguments, you can keep track of the arguments within the scope of your span. Trace arguments are disabled by default; however, they can easily be enabled by adding trace_args = True in the @Traceabledecorator.

Configuring Trace Return Values

By configuring trace return values, you can monitor all values within the scope of your span. Trace return values are disabled by default; however, they can easily be enabled by adding trace_return_value = True in the @Traceable decorator.

Configuring Trace Errors

By configuring trace errors, you can monitor any errors that are created under the scope of your span. Trace errors are disabled by default; however, they can easily be enabled by adding trace_error = True in the @Traceable decorator.

​

​