From c2dfbcc3c3de1c32de516ec4268a602cb42e0694 Mon Sep 17 00:00:00 2001 From: saber solooki Date: Wed, 6 Nov 2024 18:10:35 +0100 Subject: [PATCH 1/6] Fix(Arq): fix integration with Worker settings as a dict (#3742) --- sentry_sdk/integrations/arq.py | 11 +++ tests/integrations/arq/test_arq.py | 113 +++++++++++++++++++++++++---- 2 files changed, 110 insertions(+), 14 deletions(-) diff --git a/sentry_sdk/integrations/arq.py b/sentry_sdk/integrations/arq.py index 4640204725..d568714fe2 100644 --- a/sentry_sdk/integrations/arq.py +++ b/sentry_sdk/integrations/arq.py @@ -198,6 +198,17 @@ def _sentry_create_worker(*args, **kwargs): # type: (*Any, **Any) -> Worker settings_cls = args[0] + if isinstance(settings_cls, dict): + if "functions" in settings_cls: + settings_cls["functions"] = [ + _get_arq_function(func) for func in settings_cls["functions"] + ] + if "cron_jobs" in settings_cls: + settings_cls["cron_jobs"] = [ + _get_arq_cron_job(cron_job) + for cron_job in settings_cls["cron_jobs"] + ] + if hasattr(settings_cls, "functions"): settings_cls.functions = [ _get_arq_function(func) for func in settings_cls.functions diff --git a/tests/integrations/arq/test_arq.py b/tests/integrations/arq/test_arq.py index cd4cad67b8..e74395e26c 100644 --- a/tests/integrations/arq/test_arq.py +++ b/tests/integrations/arq/test_arq.py @@ -83,14 +83,65 @@ class WorkerSettings: return inner +@pytest.fixture +def init_arq_with_dict_settings(sentry_init): + def inner( + cls_functions=None, + cls_cron_jobs=None, + kw_functions=None, + kw_cron_jobs=None, + allow_abort_jobs_=False, + ): + cls_functions = cls_functions or [] + cls_cron_jobs = cls_cron_jobs or [] + + kwargs = {} + if kw_functions is not None: + kwargs["functions"] = kw_functions + if kw_cron_jobs is not None: + kwargs["cron_jobs"] = kw_cron_jobs + + sentry_init( + integrations=[ArqIntegration()], + traces_sample_rate=1.0, + send_default_pii=True, + ) + + server = FakeRedis() + pool = ArqRedis(pool_or_conn=server.connection_pool) + + worker_settings = { + "functions": cls_functions, + "cron_jobs": cls_cron_jobs, + "redis_pool": pool, + "allow_abort_jobs": allow_abort_jobs_, + } + + if not worker_settings["functions"]: + del worker_settings["functions"] + if not worker_settings["cron_jobs"]: + del worker_settings["cron_jobs"] + + worker = arq.worker.create_worker(worker_settings, **kwargs) + + return pool, worker + + return inner + + @pytest.mark.asyncio -async def test_job_result(init_arq): +@pytest.mark.parametrize( + "init_arq_settings", ["init_arq", "init_arq_with_dict_settings"] +) +async def test_job_result(init_arq_settings, request): async def increase(ctx, num): return num + 1 + init_fixture_method = request.getfixturevalue(init_arq_settings) + increase.__qualname__ = increase.__name__ - pool, worker = init_arq([increase]) + pool, worker = init_fixture_method([increase]) job = await pool.enqueue_job("increase", 3) @@ -105,14 +156,19 @@ async def increase(ctx, num): @pytest.mark.asyncio -async def test_job_retry(capture_events, init_arq): +@pytest.mark.parametrize( + "init_arq_settings", ["init_arq", "init_arq_with_dict_settings"] +) +async def test_job_retry(capture_events, init_arq_settings, request): async def retry_job(ctx): if ctx["job_try"] < 2: raise arq.worker.Retry + init_fixture_method = request.getfixturevalue(init_arq_settings) + retry_job.__qualname__ = retry_job.__name__ - pool, worker = init_arq([retry_job]) + pool, worker = init_fixture_method([retry_job]) job = await pool.enqueue_job("retry_job") @@ -139,11 +195,18 @@ async def retry_job(ctx): "source", [("cls_functions", "cls_cron_jobs"), ("kw_functions", "kw_cron_jobs")] ) @pytest.mark.parametrize("job_fails", [True, False], ids=["error", "success"]) +@pytest.mark.parametrize( + "init_arq_settings", ["init_arq", "init_arq_with_dict_settings"] +) @pytest.mark.asyncio -async def test_job_transaction(capture_events, init_arq, source, job_fails): +async def test_job_transaction( + capture_events, init_arq_settings, source, job_fails, request +): async def division(_, a, b=0): return a / b + init_fixture_method = request.getfixturevalue(init_arq_settings) + division.__qualname__ = division.__name__ cron_func = async_partial(division, a=1, b=int(not job_fails)) @@ -152,7 +215,9 @@ async def division(_, a, b=0): cron_job = cron(cron_func, minute=0, run_at_startup=True) functions_key, cron_jobs_key = source - pool, worker = init_arq(**{functions_key: [division], cron_jobs_key: [cron_job]}) + pool, worker = init_fixture_method( + **{functions_key: [division], cron_jobs_key: [cron_job]} + ) events = capture_events() @@ -213,12 +278,17 @@ async def division(_, a, b=0): @pytest.mark.parametrize("source", ["cls_functions", "kw_functions"]) +@pytest.mark.parametrize( + "init_arq_settings", ["init_arq", "init_arq_with_dict_settings"] +) @pytest.mark.asyncio -async def test_enqueue_job(capture_events, init_arq, source): +async def test_enqueue_job(capture_events, init_arq_settings, source, request): async def dummy_job(_): pass - pool, _ = init_arq(**{source: [dummy_job]}) + init_fixture_method = request.getfixturevalue(init_arq_settings) + + pool, _ = init_fixture_method(**{source: [dummy_job]}) events = capture_events() @@ -236,13 +306,18 @@ async def dummy_job(_): @pytest.mark.asyncio -async def test_execute_job_without_integration(init_arq): +@pytest.mark.parametrize( + "init_arq_settings", ["init_arq", "init_arq_with_dict_settings"] +) +async def test_execute_job_without_integration(init_arq_settings, request): async def dummy_job(_ctx): pass + init_fixture_method = request.getfixturevalue(init_arq_settings) + dummy_job.__qualname__ = dummy_job.__name__ - pool, worker = init_arq([dummy_job]) + pool, worker = init_fixture_method([dummy_job]) # remove the integration to trigger the edge case get_client().integrations.pop("arq") @@ -254,12 +329,17 @@ async def dummy_job(_ctx): @pytest.mark.parametrize("source", ["cls_functions", "kw_functions"]) +@pytest.mark.parametrize( + "init_arq_settings", ["init_arq", "init_arq_with_dict_settings"] +) @pytest.mark.asyncio -async def test_span_origin_producer(capture_events, init_arq, source): +async def test_span_origin_producer(capture_events, init_arq_settings, source, request): async def dummy_job(_): pass - pool, _ = init_arq(**{source: [dummy_job]}) + init_fixture_method = request.getfixturevalue(init_arq_settings) + + pool, _ = init_fixture_method(**{source: [dummy_job]}) events = capture_events() @@ -272,13 +352,18 @@ async def dummy_job(_): @pytest.mark.asyncio -async def test_span_origin_consumer(capture_events, init_arq): +@pytest.mark.parametrize( + "init_arq_settings", ["init_arq", "init_arq_with_dict_settings"] +) +async def test_span_origin_consumer(capture_events, init_arq_settings, request): async def job(ctx): pass + init_fixture_method = request.getfixturevalue(init_arq_settings) + job.__qualname__ = job.__name__ - pool, worker = init_arq([job]) + pool, worker = init_fixture_method([job]) job = await pool.enqueue_job("retry_job") From 200d0cdde8eed2caa89b91db8b17baabe983d2de Mon Sep 17 00:00:00 2001 From: Guilherme Martins Crocetti <24530683+gmcrocetti@users.noreply.github.com> Date: Thu, 7 Nov 2024 11:19:03 -0300 Subject: [PATCH 2/6] Handle parameter `stack_info` for the `LoggingIntegration` Add capability for the logging integration to use the parameter 'stack_info' (added in Python 3.2). When set to True the stack trace will be retrieved and properly handled. Fixes #2804 --- sentry_sdk/integrations/logging.py | 2 +- tests/integrations/logging/test_logging.py | 11 +++++++++-- 2 files changed, 10 insertions(+), 3 deletions(-) diff --git a/sentry_sdk/integrations/logging.py b/sentry_sdk/integrations/logging.py index 5d23440ad1..b792510d6c 100644 --- a/sentry_sdk/integrations/logging.py +++ b/sentry_sdk/integrations/logging.py @@ -202,7 +202,7 @@ def _emit(self, record): client_options=client_options, mechanism={"type": "logging", "handled": True}, ) - elif record.exc_info and record.exc_info[0] is None: + elif (record.exc_info and record.exc_info[0] is None) or record.stack_info: event = {} hint = {} with capture_internal_exceptions(): diff --git a/tests/integrations/logging/test_logging.py b/tests/integrations/logging/test_logging.py index 02eb26a04d..8c325bc86c 100644 --- a/tests/integrations/logging/test_logging.py +++ b/tests/integrations/logging/test_logging.py @@ -77,11 +77,18 @@ def test_logging_extra_data_integer_keys(sentry_init, capture_events): assert event["extra"] == {"1": 1} -def test_logging_stack(sentry_init, capture_events): +@pytest.mark.parametrize( + "enable_stack_trace_kwarg", + ( + pytest.param({"exc_info": True}, id="exc_info"), + pytest.param({"stack_info": True}, id="stack_info"), + ), +) +def test_logging_stack_trace(sentry_init, capture_events, enable_stack_trace_kwarg): sentry_init(integrations=[LoggingIntegration()], default_integrations=False) events = capture_events() - logger.error("first", exc_info=True) + logger.error("first", **enable_stack_trace_kwarg) logger.error("second") ( From d42422674379afd90ac5039e4fbac13281178ff2 Mon Sep 17 00:00:00 2001 From: Daniel Szoke <7881302+szokeasaurusrex@users.noreply.github.com> Date: Tue, 12 Nov 2024 09:16:11 +0100 Subject: [PATCH 3/6] ref(init): Deprecate `sentry_sdk.init` context manager (#3729) It is possible to use the return value of `sentry_sdk.init` as a context manager; however, this functionality has not been maintained for a long time, and it does not seem to be documented anywhere. So, we are deprecating this functionality, and we will remove it in the next major release. Closes #3282 --- sentry_sdk/_init_implementation.py | 21 +++++++++++++++++++++ tests/test_api.py | 17 +++++++++++++++++ 2 files changed, 38 insertions(+) diff --git a/sentry_sdk/_init_implementation.py b/sentry_sdk/_init_implementation.py index 256a69ee83..eb02b3d11e 100644 --- a/sentry_sdk/_init_implementation.py +++ b/sentry_sdk/_init_implementation.py @@ -1,3 +1,5 @@ +import warnings + from typing import TYPE_CHECKING import sentry_sdk @@ -9,16 +11,35 @@ class _InitGuard: + _CONTEXT_MANAGER_DEPRECATION_WARNING_MESSAGE = ( + "Using the return value of sentry_sdk.init as a context manager " + "and manually calling the __enter__ and __exit__ methods on the " + "return value are deprecated. We are no longer maintaining this " + "functionality, and we will remove it in the next major release." + ) + def __init__(self, client): # type: (sentry_sdk.Client) -> None self._client = client def __enter__(self): # type: () -> _InitGuard + warnings.warn( + self._CONTEXT_MANAGER_DEPRECATION_WARNING_MESSAGE, + stacklevel=2, + category=DeprecationWarning, + ) + return self def __exit__(self, exc_type, exc_value, tb): # type: (Any, Any, Any) -> None + warnings.warn( + self._CONTEXT_MANAGER_DEPRECATION_WARNING_MESSAGE, + stacklevel=2, + category=DeprecationWarning, + ) + c = self._client if c is not None: c.close() diff --git a/tests/test_api.py b/tests/test_api.py index ae194af7fd..3b2a9c8fb7 100644 --- a/tests/test_api.py +++ b/tests/test_api.py @@ -1,6 +1,7 @@ import pytest from unittest import mock +import sentry_sdk from sentry_sdk import ( capture_exception, continue_trace, @@ -195,3 +196,19 @@ def test_push_scope_deprecation(): with pytest.warns(DeprecationWarning): with push_scope(): ... + + +def test_init_context_manager_deprecation(): + with pytest.warns(DeprecationWarning): + with sentry_sdk.init(): + ... + + +def test_init_enter_deprecation(): + with pytest.warns(DeprecationWarning): + sentry_sdk.init().__enter__() + + +def test_init_exit_deprecation(): + with pytest.warns(DeprecationWarning): + sentry_sdk.init().__exit__(None, None, None) From 417be9ffe5e2c72e459646dc7ec14399f78c015e Mon Sep 17 00:00:00 2001 From: Burak Yigit Kaya Date: Tue, 12 Nov 2024 13:28:51 +0000 Subject: [PATCH 4/6] feat(spotlight): Inject Spotlight button on Django (#3751) This patch expands the `SpotlightMiddleware` for Django and injects the Spotlight button to all HTML responses when Spotlight is enabled and running. It requires Spotlight 2.6.0 to work this way. Ref: getsentry/spotlight#543 --- sentry_sdk/spotlight.py | 159 ++++++++++++++++++++++++++++++++-------- 1 file changed, 130 insertions(+), 29 deletions(-) diff --git a/sentry_sdk/spotlight.py b/sentry_sdk/spotlight.py index e7e90f9822..806ba5a09e 100644 --- a/sentry_sdk/spotlight.py +++ b/sentry_sdk/spotlight.py @@ -5,8 +5,9 @@ import urllib.request import urllib.error import urllib3 +import sys -from itertools import chain +from itertools import chain, product from typing import TYPE_CHECKING @@ -15,11 +16,19 @@ from typing import Callable from typing import Dict from typing import Optional + from typing import Self -from sentry_sdk.utils import logger, env_to_bool, capture_internal_exceptions +from sentry_sdk.utils import ( + logger as sentry_logger, + env_to_bool, + capture_internal_exceptions, +) from sentry_sdk.envelope import Envelope +logger = logging.getLogger("spotlight") + + DEFAULT_SPOTLIGHT_URL = "http://localhost:8969/stream" DJANGO_SPOTLIGHT_MIDDLEWARE_PATH = "sentry_sdk.spotlight.SpotlightMiddleware" @@ -34,7 +43,7 @@ def __init__(self, url): def capture_envelope(self, envelope): # type: (Envelope) -> None if self.tries > 3: - logger.warning( + sentry_logger.warning( "Too many errors sending to Spotlight, stop sending events there." ) return @@ -52,50 +61,137 @@ def capture_envelope(self, envelope): req.close() except Exception as e: self.tries += 1 - logger.warning(str(e)) + sentry_logger.warning(str(e)) try: - from django.http import HttpResponseServerError + from django.utils.deprecation import MiddlewareMixin + from django.http import HttpResponseServerError, HttpResponse, HttpRequest from django.conf import settings - class SpotlightMiddleware: - def __init__(self, get_response): - # type: (Any, Callable[..., Any]) -> None - self.get_response = get_response - - def __call__(self, request): - # type: (Any, Any) -> Any - return self.get_response(request) + SPOTLIGHT_JS_ENTRY_PATH = "/assets/main.js" + SPOTLIGHT_JS_SNIPPET_PATTERN = ( + '' + ) + SPOTLIGHT_ERROR_PAGE_SNIPPET = ( + '\n' + '\n' + ) + CHARSET_PREFIX = "charset=" + BODY_TAG_NAME = "body" + BODY_CLOSE_TAG_POSSIBILITIES = tuple( + "".format("".join(chars)) + for chars in product(*zip(BODY_TAG_NAME.upper(), BODY_TAG_NAME.lower())) + ) + + class SpotlightMiddleware(MiddlewareMixin): # type: ignore[misc] + _spotlight_script = None # type: Optional[str] - def process_exception(self, _request, exception): - # type: (Any, Any, Exception) -> Optional[HttpResponseServerError] - if not settings.DEBUG: - return None + def __init__(self, get_response): + # type: (Self, Callable[..., HttpResponse]) -> None + super().__init__(get_response) import sentry_sdk.api - spotlight_client = sentry_sdk.api.get_client().spotlight + self.sentry_sdk = sentry_sdk.api + + spotlight_client = self.sentry_sdk.get_client().spotlight if spotlight_client is None: + sentry_logger.warning( + "Cannot find Spotlight client from SpotlightMiddleware, disabling the middleware." + ) return None - # Spotlight URL has a trailing `/stream` part at the end so split it off - spotlight_url = spotlight_client.url.rsplit("/", 1)[0] + self._spotlight_url = urllib.parse.urljoin(spotlight_client.url, "../") + + @property + def spotlight_script(self): + # type: (Self) -> Optional[str] + if self._spotlight_script is None: + try: + spotlight_js_url = urllib.parse.urljoin( + self._spotlight_url, SPOTLIGHT_JS_ENTRY_PATH + ) + req = urllib.request.Request( + spotlight_js_url, + method="HEAD", + ) + urllib.request.urlopen(req) + self._spotlight_script = SPOTLIGHT_JS_SNIPPET_PATTERN.format( + spotlight_js_url + ) + except urllib.error.URLError as err: + sentry_logger.debug( + "Cannot get Spotlight JS to inject at %s. SpotlightMiddleware will not be very useful.", + spotlight_js_url, + exc_info=err, + ) + + return self._spotlight_script + + def process_response(self, _request, response): + # type: (Self, HttpRequest, HttpResponse) -> Optional[HttpResponse] + content_type_header = tuple( + p.strip() + for p in response.headers.get("Content-Type", "").lower().split(";") + ) + content_type = content_type_header[0] + if len(content_type_header) > 1 and content_type_header[1].startswith( + CHARSET_PREFIX + ): + encoding = content_type_header[1][len(CHARSET_PREFIX) :] + else: + encoding = "utf-8" + + if ( + self.spotlight_script is not None + and not response.streaming + and content_type == "text/html" + ): + content_length = len(response.content) + injection = self.spotlight_script.encode(encoding) + injection_site = next( + ( + idx + for idx in ( + response.content.rfind(body_variant.encode(encoding)) + for body_variant in BODY_CLOSE_TAG_POSSIBILITIES + ) + if idx > -1 + ), + content_length, + ) + + # This approach works even when we don't have a `` tag + response.content = ( + response.content[:injection_site] + + injection + + response.content[injection_site:] + ) + + if response.has_header("Content-Length"): + response.headers["Content-Length"] = content_length + len(injection) + + return response + + def process_exception(self, _request, exception): + # type: (Self, HttpRequest, Exception) -> Optional[HttpResponseServerError] + if not settings.DEBUG: + return None try: - spotlight = urllib.request.urlopen(spotlight_url).read().decode("utf-8") + spotlight = ( + urllib.request.urlopen(self._spotlight_url).read().decode("utf-8") + ) except urllib.error.URLError: return None else: - event_id = sentry_sdk.api.capture_exception(exception) + event_id = self.sentry_sdk.capture_exception(exception) return HttpResponseServerError( spotlight.replace( "", - ( - f'' - ''.format( - event_id=event_id - ) + SPOTLIGHT_ERROR_PAGE_SNIPPET.format( + spotlight_url=self._spotlight_url, event_id=event_id ), ) ) @@ -106,6 +202,10 @@ def process_exception(self, _request, exception): def setup_spotlight(options): # type: (Dict[str, Any]) -> Optional[SpotlightClient] + _handler = logging.StreamHandler(sys.stderr) + _handler.setFormatter(logging.Formatter(" [spotlight] %(levelname)s: %(message)s")) + logger.addHandler(_handler) + logger.setLevel(logging.INFO) url = options.get("spotlight") @@ -119,6 +219,7 @@ def setup_spotlight(options): settings is not None and settings.DEBUG and env_to_bool(os.environ.get("SENTRY_SPOTLIGHT_ON_ERROR", "1")) + and env_to_bool(os.environ.get("SENTRY_SPOTLIGHT_MIDDLEWARE", "1")) ): with capture_internal_exceptions(): middleware = settings.MIDDLEWARE @@ -126,9 +227,9 @@ def setup_spotlight(options): settings.MIDDLEWARE = type(middleware)( chain(middleware, (DJANGO_SPOTLIGHT_MIDDLEWARE_PATH,)) ) - logging.info("Enabled Spotlight integration for Django") + logger.info("Enabled Spotlight integration for Django") client = SpotlightClient(url) - logging.info("Enabled Spotlight at %s", url) + logger.info("Enabled Spotlight using sidecar at %s", url) return client From c2361a32d58eb38465e41c967788cae991a4e510 Mon Sep 17 00:00:00 2001 From: Anton Pirker Date: Wed, 13 Nov 2024 13:50:01 +0100 Subject: [PATCH 5/6] Fix aws lambda tests (by reducing event size) (#3770) Our AWS Lambda tests rely on outputting our events as JSON to stdout and parsing this output. AWS Lambda limits the amount of stdout it returns. So by reducing the size of the events we can fix the tests, that where broken by printing to much data to stdout so the output is truncated and can not be parsed into actual JSON structures again. --- tests/integrations/aws_lambda/test_aws.py | 22 +++++++++++----------- 1 file changed, 11 insertions(+), 11 deletions(-) diff --git a/tests/integrations/aws_lambda/test_aws.py b/tests/integrations/aws_lambda/test_aws.py index 75dc930da5..e229812336 100644 --- a/tests/integrations/aws_lambda/test_aws.py +++ b/tests/integrations/aws_lambda/test_aws.py @@ -98,7 +98,7 @@ def truncate_data(data): elif key == "cloudwatch logs": for cloudwatch_key in data["extra"]["cloudwatch logs"].keys(): if cloudwatch_key in ["url", "log_group", "log_stream"]: - cleaned_data["extra"].setdefault("cloudwatch logs", {})[cloudwatch_key] = data["extra"]["cloudwatch logs"][cloudwatch_key] + cleaned_data["extra"].setdefault("cloudwatch logs", {})[cloudwatch_key] = data["extra"]["cloudwatch logs"][cloudwatch_key].split("=")[0] if data.get("level") is not None: cleaned_data["level"] = data.get("level") @@ -228,7 +228,7 @@ def test_handler(event, context): assert event["extra"]["lambda"]["function_name"].startswith("test_") logs_url = event["extra"]["cloudwatch logs"]["url"] - assert logs_url.startswith("https://console.aws.amazon.com/cloudwatch/home?region=") + assert logs_url.startswith("https://console.aws.amazon.com/cloudwatch/home?region") assert not re.search("(=;|=$)", logs_url) assert event["extra"]["cloudwatch logs"]["log_group"].startswith( "/aws/lambda/test_" @@ -370,7 +370,7 @@ def test_handler(event, context): assert event["extra"]["lambda"]["function_name"].startswith("test_") logs_url = event["extra"]["cloudwatch logs"]["url"] - assert logs_url.startswith("https://console.aws.amazon.com/cloudwatch/home?region=") + assert logs_url.startswith("https://console.aws.amazon.com/cloudwatch/home?region") assert not re.search("(=;|=$)", logs_url) assert event["extra"]["cloudwatch logs"]["log_group"].startswith( "/aws/lambda/test_" @@ -462,11 +462,11 @@ def test_handler(event, context): "X-Forwarded-Proto": "https" }, "httpMethod": "GET", - "path": "/path1", + "path": "/1", "queryStringParameters": { - "done": "false" + "done": "f" }, - "dog": "Maisey" + "d": "D1" }, { "headers": { @@ -474,11 +474,11 @@ def test_handler(event, context): "X-Forwarded-Proto": "http" }, "httpMethod": "POST", - "path": "/path2", + "path": "/2", "queryStringParameters": { - "done": "true" + "done": "t" }, - "dog": "Charlie" + "d": "D2" } ] """, @@ -538,9 +538,9 @@ def test_handler(event, context): request_data = { "headers": {"Host": "x1.io", "X-Forwarded-Proto": "https"}, "method": "GET", - "url": "https://x1.io/path1", + "url": "https://x1.io/1", "query_string": { - "done": "false", + "done": "f", }, } else: From 4bec4a4729b64525ef55947fd4042e0d62ef72cc Mon Sep 17 00:00:00 2001 From: matt-codecov <137832199+matt-codecov@users.noreply.github.com> Date: Wed, 13 Nov 2024 05:30:58 -0800 Subject: [PATCH 6/6] feat: introduce rust_tracing integration (#3717) Introduce a new integration that allows traces to descend into code in Rust native extensions by hooking into Rust's popular `tracing` framework. it relies on the Rust native extension using [`pyo3-python-tracing-subscriber`](https://crates.io/crates/pyo3-python-tracing-subscriber), a crate i recently published under Sentry, to expose a way for the Python SDK to hook into `tracing`. in this screenshot, the transaction was started in Python but the rest of the span tree reflects the structure and performance of a naive fibonacci generator in Rust: https://github.com/user-attachments/assets/ae2caff6-1842-45d0-a604-2f3b6305f330 --------- Co-authored-by: Anton Pirker --- sentry_sdk/integrations/rust_tracing.py | 274 +++++++++++ tests/integrations/rust_tracing/__init__.py | 0 .../rust_tracing/test_rust_tracing.py | 450 ++++++++++++++++++ 3 files changed, 724 insertions(+) create mode 100644 sentry_sdk/integrations/rust_tracing.py create mode 100644 tests/integrations/rust_tracing/__init__.py create mode 100644 tests/integrations/rust_tracing/test_rust_tracing.py diff --git a/sentry_sdk/integrations/rust_tracing.py b/sentry_sdk/integrations/rust_tracing.py new file mode 100644 index 0000000000..121bf082b8 --- /dev/null +++ b/sentry_sdk/integrations/rust_tracing.py @@ -0,0 +1,274 @@ +""" +This integration ingests tracing data from native extensions written in Rust. + +Using it requires additional setup on the Rust side to accept a +`RustTracingLayer` Python object and register it with the `tracing-subscriber` +using an adapter from the `pyo3-python-tracing-subscriber` crate. For example: +```rust +#[pyfunction] +pub fn initialize_tracing(py_impl: Bound<'_, PyAny>) { + tracing_subscriber::registry() + .with(pyo3_python_tracing_subscriber::PythonCallbackLayerBridge::new(py_impl)) + .init(); +} +``` + +Usage in Python would then look like: +``` +sentry_sdk.init( + dsn=sentry_dsn, + integrations=[ + RustTracingIntegration( + "demo_rust_extension", + demo_rust_extension.initialize_tracing, + event_type_mapping=event_type_mapping, + ) + ], +) +``` + +Each native extension requires its own integration. +""" + +import json +from enum import Enum, auto +from typing import Any, Callable, Dict, Tuple, Optional + +import sentry_sdk +from sentry_sdk.integrations import Integration +from sentry_sdk.scope import should_send_default_pii +from sentry_sdk.tracing import Span as SentrySpan +from sentry_sdk.utils import SENSITIVE_DATA_SUBSTITUTE + +TraceState = Optional[Tuple[Optional[SentrySpan], SentrySpan]] + + +class RustTracingLevel(Enum): + Trace: str = "TRACE" + Debug: str = "DEBUG" + Info: str = "INFO" + Warn: str = "WARN" + Error: str = "ERROR" + + +class EventTypeMapping(Enum): + Ignore = auto() + Exc = auto() + Breadcrumb = auto() + Event = auto() + + +def tracing_level_to_sentry_level(level): + # type: (str) -> sentry_sdk._types.LogLevelStr + level = RustTracingLevel(level) + if level in (RustTracingLevel.Trace, RustTracingLevel.Debug): + return "debug" + elif level == RustTracingLevel.Info: + return "info" + elif level == RustTracingLevel.Warn: + return "warning" + elif level == RustTracingLevel.Error: + return "error" + else: + # Better this than crashing + return "info" + + +def extract_contexts(event: Dict[str, Any]) -> Dict[str, Any]: + metadata = event.get("metadata", {}) + contexts = {} + + location = {} + for field in ["module_path", "file", "line"]: + if field in metadata: + location[field] = metadata[field] + if len(location) > 0: + contexts["rust_tracing_location"] = location + + fields = {} + for field in metadata.get("fields", []): + fields[field] = event.get(field) + if len(fields) > 0: + contexts["rust_tracing_fields"] = fields + + return contexts + + +def process_event(event: Dict[str, Any]) -> None: + metadata = event.get("metadata", {}) + + logger = metadata.get("target") + level = tracing_level_to_sentry_level(metadata.get("level")) + message = event.get("message") # type: sentry_sdk._types.Any + contexts = extract_contexts(event) + + sentry_event = { + "logger": logger, + "level": level, + "message": message, + "contexts": contexts, + } # type: sentry_sdk._types.Event + + sentry_sdk.capture_event(sentry_event) + + +def process_exception(event: Dict[str, Any]) -> None: + process_event(event) + + +def process_breadcrumb(event: Dict[str, Any]) -> None: + level = tracing_level_to_sentry_level(event.get("metadata", {}).get("level")) + message = event.get("message") + + sentry_sdk.add_breadcrumb(level=level, message=message) + + +def default_span_filter(metadata: Dict[str, Any]) -> bool: + return RustTracingLevel(metadata.get("level")) in ( + RustTracingLevel.Error, + RustTracingLevel.Warn, + RustTracingLevel.Info, + ) + + +def default_event_type_mapping(metadata: Dict[str, Any]) -> EventTypeMapping: + level = RustTracingLevel(metadata.get("level")) + if level == RustTracingLevel.Error: + return EventTypeMapping.Exc + elif level in (RustTracingLevel.Warn, RustTracingLevel.Info): + return EventTypeMapping.Breadcrumb + elif level in (RustTracingLevel.Debug, RustTracingLevel.Trace): + return EventTypeMapping.Ignore + else: + return EventTypeMapping.Ignore + + +class RustTracingLayer: + def __init__( + self, + origin: str, + event_type_mapping: Callable[ + [Dict[str, Any]], EventTypeMapping + ] = default_event_type_mapping, + span_filter: Callable[[Dict[str, Any]], bool] = default_span_filter, + send_sensitive_data: Optional[bool] = None, + ): + self.origin = origin + self.event_type_mapping = event_type_mapping + self.span_filter = span_filter + self.send_sensitive_data = send_sensitive_data + + def on_event(self, event: str, _span_state: TraceState) -> None: + deserialized_event = json.loads(event) + metadata = deserialized_event.get("metadata", {}) + + event_type = self.event_type_mapping(metadata) + if event_type == EventTypeMapping.Ignore: + return + elif event_type == EventTypeMapping.Exc: + process_exception(deserialized_event) + elif event_type == EventTypeMapping.Breadcrumb: + process_breadcrumb(deserialized_event) + elif event_type == EventTypeMapping.Event: + process_event(deserialized_event) + + def on_new_span(self, attrs: str, span_id: str) -> TraceState: + attrs = json.loads(attrs) + metadata = attrs.get("metadata", {}) + + if not self.span_filter(metadata): + return None + + module_path = metadata.get("module_path") + name = metadata.get("name") + message = attrs.get("message") + + if message is not None: + sentry_span_name = message + elif module_path is not None and name is not None: + sentry_span_name = f"{module_path}::{name}" # noqa: E231 + elif name is not None: + sentry_span_name = name + else: + sentry_span_name = "" + + kwargs = { + "op": "function", + "name": sentry_span_name, + "origin": self.origin, + } + + scope = sentry_sdk.get_current_scope() + parent_sentry_span = scope.span + if parent_sentry_span: + sentry_span = parent_sentry_span.start_child(**kwargs) + else: + sentry_span = scope.start_span(**kwargs) + + fields = metadata.get("fields", []) + for field in fields: + sentry_span.set_data(field, attrs.get(field)) + + scope.span = sentry_span + return (parent_sentry_span, sentry_span) + + def on_close(self, span_id: str, span_state: TraceState) -> None: + if span_state is None: + return + + parent_sentry_span, sentry_span = span_state + sentry_span.finish() + sentry_sdk.get_current_scope().span = parent_sentry_span + + def on_record(self, span_id: str, values: str, span_state: TraceState) -> None: + if span_state is None: + return + _parent_sentry_span, sentry_span = span_state + + send_sensitive_data = ( + should_send_default_pii() + if self.send_sensitive_data is None + else self.send_sensitive_data + ) + + deserialized_values = json.loads(values) + for key, value in deserialized_values.items(): + if send_sensitive_data: + sentry_span.set_data(key, value) + else: + sentry_span.set_data(key, SENSITIVE_DATA_SUBSTITUTE) + + +class RustTracingIntegration(Integration): + """ + Ingests tracing data from a Rust native extension's `tracing` instrumentation. + + If a project uses more than one Rust native extension, each one will need + its own instance of `RustTracingIntegration` with an initializer function + specific to that extension. + + Since all of the setup for this integration requires instance-specific state + which is not available in `setup_once()`, setup instead happens in `__init__()`. + """ + + def __init__( + self, + identifier: str, + initializer: Callable[[RustTracingLayer], None], + event_type_mapping: Callable[ + [Dict[str, Any]], EventTypeMapping + ] = default_event_type_mapping, + span_filter: Callable[[Dict[str, Any]], bool] = default_span_filter, + send_sensitive_data: Optional[bool] = None, + ): + self.identifier = identifier + origin = f"auto.function.rust_tracing.{identifier}" + self.tracing_layer = RustTracingLayer( + origin, event_type_mapping, span_filter, send_sensitive_data + ) + + initializer(self.tracing_layer) + + @staticmethod + def setup_once() -> None: + pass diff --git a/tests/integrations/rust_tracing/__init__.py b/tests/integrations/rust_tracing/__init__.py new file mode 100644 index 0000000000..e69de29bb2 diff --git a/tests/integrations/rust_tracing/test_rust_tracing.py b/tests/integrations/rust_tracing/test_rust_tracing.py new file mode 100644 index 0000000000..b1fad1a7f7 --- /dev/null +++ b/tests/integrations/rust_tracing/test_rust_tracing.py @@ -0,0 +1,450 @@ +import pytest + +from string import Template +from typing import Dict + +import sentry_sdk +from sentry_sdk.integrations.rust_tracing import ( + RustTracingIntegration, + RustTracingLayer, + RustTracingLevel, + EventTypeMapping, +) +from sentry_sdk import start_transaction, capture_message + + +def _test_event_type_mapping(metadata: Dict[str, object]) -> EventTypeMapping: + level = RustTracingLevel(metadata.get("level")) + if level == RustTracingLevel.Error: + return EventTypeMapping.Exc + elif level in (RustTracingLevel.Warn, RustTracingLevel.Info): + return EventTypeMapping.Breadcrumb + elif level == RustTracingLevel.Debug: + return EventTypeMapping.Event + elif level == RustTracingLevel.Trace: + return EventTypeMapping.Ignore + else: + return EventTypeMapping.Ignore + + +class FakeRustTracing: + # Parameters: `level`, `index` + span_template = Template( + """{"index":$index,"is_root":false,"metadata":{"fields":["index","use_memoized","version"],"file":"src/lib.rs","is_event":false,"is_span":true,"level":"$level","line":40,"module_path":"_bindings","name":"fibonacci","target":"_bindings"},"parent":null,"use_memoized":true}""" + ) + + # Parameters: `level`, `index` + event_template = Template( + """{"message":"Getting the ${index}th fibonacci number","metadata":{"fields":["message"],"file":"src/lib.rs","is_event":true,"is_span":false,"level":"$level","line":23,"module_path":"_bindings","name":"event src/lib.rs:23","target":"_bindings"}}""" + ) + + def __init__(self): + self.spans = {} + + def set_layer_impl(self, layer: RustTracingLayer): + self.layer = layer + + def new_span(self, level: RustTracingLevel, span_id: int, index_arg: int = 10): + span_attrs = self.span_template.substitute(level=level.value, index=index_arg) + state = self.layer.on_new_span(span_attrs, str(span_id)) + self.spans[span_id] = state + + def close_span(self, span_id: int): + state = self.spans.pop(span_id) + self.layer.on_close(str(span_id), state) + + def event(self, level: RustTracingLevel, span_id: int, index_arg: int = 10): + event = self.event_template.substitute(level=level.value, index=index_arg) + state = self.spans[span_id] + self.layer.on_event(event, state) + + def record(self, span_id: int): + state = self.spans[span_id] + self.layer.on_record(str(span_id), """{"version": "memoized"}""", state) + + +def test_on_new_span_on_close(sentry_init, capture_events): + rust_tracing = FakeRustTracing() + integration = RustTracingIntegration( + "test_on_new_span_on_close", rust_tracing.set_layer_impl + ) + sentry_init(integrations=[integration], traces_sample_rate=1.0) + + events = capture_events() + with start_transaction(): + rust_tracing.new_span(RustTracingLevel.Info, 3) + + sentry_first_rust_span = sentry_sdk.get_current_span() + _, rust_first_rust_span = rust_tracing.spans[3] + + assert sentry_first_rust_span == rust_first_rust_span + + rust_tracing.close_span(3) + assert sentry_sdk.get_current_span() != sentry_first_rust_span + + (event,) = events + assert len(event["spans"]) == 1 + + # Ensure the span metadata is wired up + span = event["spans"][0] + assert span["op"] == "function" + assert span["origin"] == "auto.function.rust_tracing.test_on_new_span_on_close" + assert span["description"] == "_bindings::fibonacci" + + # Ensure the span was opened/closed appropriately + assert span["start_timestamp"] is not None + assert span["timestamp"] is not None + + # Ensure the extra data from Rust is hooked up + data = span["data"] + assert data["use_memoized"] + assert data["index"] == 10 + assert data["version"] is None + + +def test_nested_on_new_span_on_close(sentry_init, capture_events): + rust_tracing = FakeRustTracing() + integration = RustTracingIntegration( + "test_nested_on_new_span_on_close", rust_tracing.set_layer_impl + ) + sentry_init(integrations=[integration], traces_sample_rate=1.0) + + events = capture_events() + with start_transaction(): + original_sentry_span = sentry_sdk.get_current_span() + + rust_tracing.new_span(RustTracingLevel.Info, 3, index_arg=10) + sentry_first_rust_span = sentry_sdk.get_current_span() + _, rust_first_rust_span = rust_tracing.spans[3] + + # Use a different `index_arg` value for the inner span to help + # distinguish the two at the end of the test + rust_tracing.new_span(RustTracingLevel.Info, 5, index_arg=9) + sentry_second_rust_span = sentry_sdk.get_current_span() + rust_parent_span, rust_second_rust_span = rust_tracing.spans[5] + + assert rust_second_rust_span == sentry_second_rust_span + assert rust_parent_span == sentry_first_rust_span + assert rust_parent_span == rust_first_rust_span + assert rust_parent_span != rust_second_rust_span + + rust_tracing.close_span(5) + + # Ensure the current sentry span was moved back to the parent + sentry_span_after_close = sentry_sdk.get_current_span() + assert sentry_span_after_close == sentry_first_rust_span + + rust_tracing.close_span(3) + + assert sentry_sdk.get_current_span() == original_sentry_span + + (event,) = events + assert len(event["spans"]) == 2 + + # Ensure the span metadata is wired up for all spans + first_span, second_span = event["spans"] + assert first_span["op"] == "function" + assert ( + first_span["origin"] + == "auto.function.rust_tracing.test_nested_on_new_span_on_close" + ) + assert first_span["description"] == "_bindings::fibonacci" + assert second_span["op"] == "function" + assert ( + second_span["origin"] + == "auto.function.rust_tracing.test_nested_on_new_span_on_close" + ) + assert second_span["description"] == "_bindings::fibonacci" + + # Ensure the spans were opened/closed appropriately + assert first_span["start_timestamp"] is not None + assert first_span["timestamp"] is not None + assert second_span["start_timestamp"] is not None + assert second_span["timestamp"] is not None + + # Ensure the extra data from Rust is hooked up in both spans + first_span_data = first_span["data"] + assert first_span_data["use_memoized"] + assert first_span_data["index"] == 10 + assert first_span_data["version"] is None + + second_span_data = second_span["data"] + assert second_span_data["use_memoized"] + assert second_span_data["index"] == 9 + assert second_span_data["version"] is None + + +def test_on_new_span_without_transaction(sentry_init): + rust_tracing = FakeRustTracing() + integration = RustTracingIntegration( + "test_on_new_span_without_transaction", rust_tracing.set_layer_impl + ) + sentry_init(integrations=[integration], traces_sample_rate=1.0) + + assert sentry_sdk.get_current_span() is None + + # Should still create a span hierarchy, it just will not be under a txn + rust_tracing.new_span(RustTracingLevel.Info, 3) + current_span = sentry_sdk.get_current_span() + assert current_span is not None + assert current_span.containing_transaction is None + + +def test_on_event_exception(sentry_init, capture_events): + rust_tracing = FakeRustTracing() + integration = RustTracingIntegration( + "test_on_event_exception", + rust_tracing.set_layer_impl, + event_type_mapping=_test_event_type_mapping, + ) + sentry_init(integrations=[integration], traces_sample_rate=1.0) + + events = capture_events() + sentry_sdk.get_isolation_scope().clear_breadcrumbs() + + with start_transaction(): + rust_tracing.new_span(RustTracingLevel.Info, 3) + + # Mapped to Exception + rust_tracing.event(RustTracingLevel.Error, 3) + + rust_tracing.close_span(3) + + assert len(events) == 2 + exc, _tx = events + assert exc["level"] == "error" + assert exc["logger"] == "_bindings" + assert exc["message"] == "Getting the 10th fibonacci number" + assert exc["breadcrumbs"]["values"] == [] + + location_context = exc["contexts"]["rust_tracing_location"] + assert location_context["module_path"] == "_bindings" + assert location_context["file"] == "src/lib.rs" + assert location_context["line"] == 23 + + field_context = exc["contexts"]["rust_tracing_fields"] + assert field_context["message"] == "Getting the 10th fibonacci number" + + +def test_on_event_breadcrumb(sentry_init, capture_events): + rust_tracing = FakeRustTracing() + integration = RustTracingIntegration( + "test_on_event_breadcrumb", + rust_tracing.set_layer_impl, + event_type_mapping=_test_event_type_mapping, + ) + sentry_init(integrations=[integration], traces_sample_rate=1.0) + + events = capture_events() + sentry_sdk.get_isolation_scope().clear_breadcrumbs() + + with start_transaction(): + rust_tracing.new_span(RustTracingLevel.Info, 3) + + # Mapped to Breadcrumb + rust_tracing.event(RustTracingLevel.Info, 3) + + rust_tracing.close_span(3) + capture_message("test message") + + assert len(events) == 2 + message, _tx = events + + breadcrumbs = message["breadcrumbs"]["values"] + assert len(breadcrumbs) == 1 + assert breadcrumbs[0]["level"] == "info" + assert breadcrumbs[0]["message"] == "Getting the 10th fibonacci number" + assert breadcrumbs[0]["type"] == "default" + + +def test_on_event_event(sentry_init, capture_events): + rust_tracing = FakeRustTracing() + integration = RustTracingIntegration( + "test_on_event_event", + rust_tracing.set_layer_impl, + event_type_mapping=_test_event_type_mapping, + ) + sentry_init(integrations=[integration], traces_sample_rate=1.0) + + events = capture_events() + sentry_sdk.get_isolation_scope().clear_breadcrumbs() + + with start_transaction(): + rust_tracing.new_span(RustTracingLevel.Info, 3) + + # Mapped to Event + rust_tracing.event(RustTracingLevel.Debug, 3) + + rust_tracing.close_span(3) + + assert len(events) == 2 + event, _tx = events + + assert event["logger"] == "_bindings" + assert event["level"] == "debug" + assert event["message"] == "Getting the 10th fibonacci number" + assert event["breadcrumbs"]["values"] == [] + + location_context = event["contexts"]["rust_tracing_location"] + assert location_context["module_path"] == "_bindings" + assert location_context["file"] == "src/lib.rs" + assert location_context["line"] == 23 + + field_context = event["contexts"]["rust_tracing_fields"] + assert field_context["message"] == "Getting the 10th fibonacci number" + + +def test_on_event_ignored(sentry_init, capture_events): + rust_tracing = FakeRustTracing() + integration = RustTracingIntegration( + "test_on_event_ignored", + rust_tracing.set_layer_impl, + event_type_mapping=_test_event_type_mapping, + ) + sentry_init(integrations=[integration], traces_sample_rate=1.0) + + events = capture_events() + sentry_sdk.get_isolation_scope().clear_breadcrumbs() + + with start_transaction(): + rust_tracing.new_span(RustTracingLevel.Info, 3) + + # Ignored + rust_tracing.event(RustTracingLevel.Trace, 3) + + rust_tracing.close_span(3) + + assert len(events) == 1 + (tx,) = events + assert tx["type"] == "transaction" + assert "message" not in tx + + +def test_span_filter(sentry_init, capture_events): + def span_filter(metadata: Dict[str, object]) -> bool: + return RustTracingLevel(metadata.get("level")) in ( + RustTracingLevel.Error, + RustTracingLevel.Warn, + RustTracingLevel.Info, + RustTracingLevel.Debug, + ) + + rust_tracing = FakeRustTracing() + integration = RustTracingIntegration( + "test_span_filter", rust_tracing.set_layer_impl, span_filter=span_filter + ) + sentry_init(integrations=[integration], traces_sample_rate=1.0) + + events = capture_events() + with start_transaction(): + original_sentry_span = sentry_sdk.get_current_span() + + # Span is not ignored + rust_tracing.new_span(RustTracingLevel.Info, 3, index_arg=10) + info_span = sentry_sdk.get_current_span() + + # Span is ignored, current span should remain the same + rust_tracing.new_span(RustTracingLevel.Trace, 5, index_arg=9) + assert sentry_sdk.get_current_span() == info_span + + # Closing the filtered span should leave the current span alone + rust_tracing.close_span(5) + assert sentry_sdk.get_current_span() == info_span + + rust_tracing.close_span(3) + assert sentry_sdk.get_current_span() == original_sentry_span + + (event,) = events + assert len(event["spans"]) == 1 + # The ignored span has index == 9 + assert event["spans"][0]["data"]["index"] == 10 + + +def test_record(sentry_init): + rust_tracing = FakeRustTracing() + integration = RustTracingIntegration( + "test_record", + initializer=rust_tracing.set_layer_impl, + send_sensitive_data=True, + ) + sentry_init(integrations=[integration], traces_sample_rate=1.0) + + with start_transaction(): + rust_tracing.new_span(RustTracingLevel.Info, 3) + + span_before_record = sentry_sdk.get_current_span().to_json() + assert span_before_record["data"]["version"] is None + + rust_tracing.record(3) + + span_after_record = sentry_sdk.get_current_span().to_json() + assert span_after_record["data"]["version"] == "memoized" + + +def test_record_in_ignored_span(sentry_init): + def span_filter(metadata: Dict[str, object]) -> bool: + # Just ignore Trace + return RustTracingLevel(metadata.get("level")) != RustTracingLevel.Trace + + rust_tracing = FakeRustTracing() + integration = RustTracingIntegration( + "test_record_in_ignored_span", + rust_tracing.set_layer_impl, + span_filter=span_filter, + ) + sentry_init(integrations=[integration], traces_sample_rate=1.0) + + with start_transaction(): + rust_tracing.new_span(RustTracingLevel.Info, 3) + + span_before_record = sentry_sdk.get_current_span().to_json() + assert span_before_record["data"]["version"] is None + + rust_tracing.new_span(RustTracingLevel.Trace, 5) + rust_tracing.record(5) + + # `on_record()` should not do anything to the current Sentry span if the associated Rust span was ignored + span_after_record = sentry_sdk.get_current_span().to_json() + assert span_after_record["data"]["version"] is None + + +@pytest.mark.parametrize( + "send_default_pii, send_sensitive_data, sensitive_data_expected", + [ + (True, True, True), + (True, False, False), + (True, None, True), + (False, True, True), + (False, False, False), + (False, None, False), + ], +) +def test_sensitive_data( + sentry_init, send_default_pii, send_sensitive_data, sensitive_data_expected +): + rust_tracing = FakeRustTracing() + integration = RustTracingIntegration( + "test_record", + initializer=rust_tracing.set_layer_impl, + send_sensitive_data=send_sensitive_data, + ) + + sentry_init( + integrations=[integration], + traces_sample_rate=1.0, + send_default_pii=send_default_pii, + ) + with start_transaction(): + rust_tracing.new_span(RustTracingLevel.Info, 3) + + span_before_record = sentry_sdk.get_current_span().to_json() + assert span_before_record["data"]["version"] is None + + rust_tracing.record(3) + + span_after_record = sentry_sdk.get_current_span().to_json() + + if sensitive_data_expected: + assert span_after_record["data"]["version"] == "memoized" + else: + assert span_after_record["data"]["version"] == "[Filtered]"