urllib instrumentation fails for local file access

[gregbuehler]

When reading local files the status code is not specified and is None. This isn't handled by the instrumentation and causes an exception.

opentelemetry-python-contrib/instrumentation/opentelemetry-instrumentation-urllib/src/opentelemetry/instrumentation/urllib/__init__.py

Lines 212 to 217 in 444e0a1

	code_ = result.getcode()
	labels[SpanAttributes.HTTP_STATUS_CODE] = str(code_)
	if span.is_recording():
	span.set_attribute(SpanAttributes.HTTP_STATUS_CODE, code_)
	span.set_status(Status(http_status_to_status_code(code_)))

[owais]

Thanks for reporting this. Could you please share the exception traceback and a reproducible code example?

[gregbuehler]

Traceback:

Traceback (most recent call last):
  File "/opt/foo/lib/python3.7/site-packages/swagger_spec_validator/common.py", line 36, in wrapper
    return method(*args, **kwargs)
  File "/opt/foo/lib/python3.7/site-packages/swagger_spec_validator/validator20.py", line 195, in validate_json
    schema, schema_path = read_resource_file(schema_path)
  File "/opt/foo/lib/python3.7/site-packages/swagger_spec_validator/common.py", line 63, in read_resource_file
    return read_file(schema_path), schema_path
  File "/opt/foo/lib/python3.7/site-packages/swagger_spec_validator/common.py", line 57, in read_file
    return read_url(get_uri_from_file_path(file_path))
  File "/opt/foo/lib/python3.7/site-packages/swagger_spec_validator/common.py", line 67, in read_url
    with contextlib.closing(urlopen(url, timeout=timeout)) as fh:
  File "/usr/lib/python3.7/urllib/request.py", line 222, in urlopen
    return opener.open(url, data, timeout)
  File "/opt/foo/lib/python3.7/site-packages/opentelemetry/instrumentation/urllib/__init__.py", line 163, in instrumented_open
    opener, request_, call_wrapped, get_or_create_headers
  File "/opt/foo/lib/python3.7/site-packages/opentelemetry/instrumentation/urllib/__init__.py", line 217, in _instrumented_open_call
    span.set_status(Status(http_status_to_status_code(code_)))
  File "/opt/foo/lib/python3.7/site-packages/opentelemetry/instrumentation/utils.py", line 49, in http_status_to_status_code
    if status < 100:
TypeError: '<' not supported between instances of 'NoneType' and 'int'

[gregbuehler]

It's not minimal, but reproduction is as easy as specifying a local file as the schema for https://github.com/Yelp/swagger_spec_validator.

[lzchen]

Is "returning a non-int status code' a common thing amongst instrumentations? If not, maybe we should treat this as a special use case and address it only for urlib.

[owais]

I think in the interest of never crashing, http_status_to_status_code should be defensive and return non-error in case it is passed invalid input.