getsentry / sentry-python

The official Python SDK for Sentry.io

Bad software is everywhere, and we're tired of it. Sentry is on a mission to help developers write better software faster, so we can get back to enjoying technology. If you want to join us Check out our open positions

Official Sentry SDK for Python

This is the official Python SDK for Sentry

---

Getting Started

Install

pip install --upgrade sentry-sdk

Configuration

import sentry_sdk

sentry_sdk.init(
    "https://[email protected]/1",

    # Set traces_sample_rate to 1.0 to capture 100%
    # of transactions for performance monitoring.
    traces_sample_rate=1.0,
)

Usage

from sentry_sdk import capture_message
capture_message("Hello World")  # Will create an event in Sentry.

raise ValueError()  # Will also create an event in Sentry.

• To learn more about how to use the SDK refer to our docs.
• Are you coming from raven-python? Use this migration guide.
• To learn about internals use the API Reference.

Integrations

(If you want to create a new integration, have a look at the Adding a new integration checklist.)

See the documentation for an up-to-date list of libraries and frameworks we support. Here are some examples:

• Django
• Flask
• FastAPI
• AIOHTTP
• SQLAlchemy
• asyncpg
• Redis
• Celery
• Apache Airflow
• Apache Spark
• asyncio
• Graphene
• Logging
• Loguru
• HTTPX
• AWS Lambda
• Google Cloud Functions

Migrating From raven-python

The old raven-python client has entered maintenance mode and was moved here.

If you're using raven-python, we recommend you to migrate to this new SDK. You can find the benefits of migrating and how to do it in our migration guide.

Contributing to the SDK

Please refer to CONTRIBUTING.md.

Getting Help/Support

If you need help setting up or configuring the Python SDK (or anything else in the Sentry universe) please head over to the Sentry Community on Discord. There is a ton of great people in our Discord community ready to help you!

Resources

• 
• 
• 
• 
• 

License

Licensed under the MIT license, see LICENSE

sentry-python's Issues

Make sure everything works with Django-Rest-Framework

We should django-rest-framework's request.data instead of trying to extract a structured body ourselves

TypeError: getresponse() got an unexpected keyword argument 'buffering'

Traceback (most recent call last):
  File "/app/.heroku/python/lib/python3.7/site-packages/botocore/vendored/requests/packages/urllib3/connectionpool.py", line 372, in _make_request
    httplib_response = conn.getresponse(buffering=True)
  File "/app/.heroku/python/lib/python3.7/site-packages/sentry_sdk/integrations/stdlib.py", line 38, in getresponse
    rv = real_getresponse(self, *args, **kwargs)
TypeError: getresponse() got an unexpected keyword argument 'buffering'

Incorrectly named option

drain_timeout -> shutdown_timeout

Split up breadcrumb and event logging handler, expose them

While migrating sentry.io to the new SDK this came up: There should be two separate handlers, one for creating events and one for creating breadcrumbs. Expose and document them as well.

Tornado support

Tracking interest at the moment. If you're interested in this please give this post a 👍 reaction. Don't post your 👍 as comment, this makes it harder to track.

(Note: The raven-python SDK supports Tornado too)

Problem with breadcrumbs

We've just started using sentry, and we are occasionally seeing this error:

AttributeError
'list' object has no attribute 'popleft'
hub.py in add_breadcrumb at line 206
        if crumb is not None:
            scope._breadcrumbs.append(crumb)
        else:
            logger.info("before breadcrumb dropped breadcrumb (%s)", original_crumb)
        while len(scope._breadcrumbs) >= client.options["max_breadcrumbs"]:
            scope._breadcrumbs.popleft()
    def push_scope(self):
        """Pushes a new layer on the scope stack. Returns a context manager
        that should be used to pop the scope again."""
        client, scope = self._stack[-1]

And, furthermore in sentry it says:

client | <sentry_sdk.client.Client object at 0x7fd665dcbcc0>
-- | --
crumb | {category: 'httplib', data: {method: 'POST', reason: 'OK', status_code: 200, url: 'https://us-west-2.queue.amazonaws.com/'}, timestamp: datetime.datetime(2018, 9, 25, 16, 20, 20, 89233), type: 'http'}
hint | {httplib_response: <http.client.HTTPResponse object at 0x7fd645fd7e80>}
kwargs | {category: 'httplib', data: {method: 'POST', reason: 'OK', status_code: 200, url: 'https://us-west-2.queue.amazonaws.com/'}, type: 'http'}
original_crumb | {category: 'httplib', data: {method: 'POST', reason: 'OK', status_code: 200, url: 'https://us-west-2.queue.amazonaws.com/'}, timestamp: datetime.datetime(2018, 9, 25, 16, 20, 20, 89233), type: 'http'}
scope | <sentry_sdk.scope.Scope object at 0x7fd665b41d68>
self | <sentry_sdk.hub.Hub object at 0x7fd665aeb320>

Any thoughts on what is going on?
Thanks!
-- Samson

Don't capture entire environ in wsgi event processor

@mitsuhiko could you look over

sentry-python/sentry_sdk/integrations/_wsgi.py

Line 205 in 98399ca

def _make_wsgi_event_processor(environ, client_options):

? I wonder if it would make sense to just write to the scope eagerly instead. Weakref on the environ dict is kind of shitty: Any middleware or web framework could do dict(environ) and the weakref is broken

Pyramid support

Pyramid was previously supported by raven. It would be great to see support for it in this client. Is this perhaps already on the roadmap?

Also if there is any documentation on how to write an integration I might be tempted to contribute.

Attribute error

AttributeError
'Identifier' object has no attribute 'postponed_alerts_counts'
ID: 8099fd1f8815496eaa1f815657918b9d
Sept. 26, 2018, 7:37:17 p.m. UTC
Exception

AttributeError: 'Identifier' object has no attribute 'postponed_alerts_counts'
File "interfaces.py", line 43, in start
self.startInternal()
File "identifier.py", line 63, in startInternal
send_time=send_time)
File "identifier.py", line 128, in match_batch
accepted = self.correct_by_past_history(match.person_guid, confident_score)
File "identifier.py", line 91, in correct_by_past_history
self.postponed_alerts_counts[guid] += 1

Figure out better place to init for Django apps

from #19

Transport does not use client's hub

The default HTTP Transport lives on its own thread and will not inherit the debug settings of the hub where the message was captured. It's basically impossible to enable debugging for threaded transports unless you make your hub the main one

[Tech Support] How to use this with django?

Where exactly am I supposed to put sentry_sdk.init()?

If I put it in settings.py, I see no error messages being reported.

The docs say nothing about dajngo + new sentry-sdk, except the fact that raven one is deprecated!

Wrong display of cyrillic characters in sentry

In Exception section in Sentry cyrillic characters do not display correctly:

Breadcrumbs type error

We're having an issue with v0.3.2 where popleft is called on a list type:

File "/srv/frontend/project/lib/python3.5/site-packages/sentry_sdk/hub.py" in add_breadcrumb
  209.             scope._breadcrumbs.popleft()

Exception Type: AttributeError at /
Exception Value: 'list' object has no attribute 'popleft'

capture WSGI-level errors

from #19

Basically if Django/Flask itself crashes (or some wsgi middleware) we won't notice.

This is slightly related to #20 because we need to wrap the WSGI app last if somebody decided to use WSGI middleware in Django.

In Flask we can wrap Flask.__call__

An unsolved question is whether we want to attach basic request info for such events, and if we're better off doing the entire request parsing in the wsgi middleware at that point.

AWS Lambda integration not working with AWS Chalice

When trying to deploy a "Chalice" serverless App, getting exception:

    File "/Users/cookie/.virtualenvs/abc/lib/python3.6/site-packages/sentry_sdk/integrations/aws_lambda.py", line 21, in install
    old_make_final_handler = lambda_bootstrap.make_final_handler
AttributeError: module '__main__' has no attribute 'make_final_handler'

sentry-python version: 0.3.9
Code in app.py:

import sentry_sdk
from sentry_sdk.integrations.aws_lambda import AwsLambdaIntegration

sentry_sdk.init(
    dsn="https://[email protected]/12345678",
    integrations=[AwsLambdaIntegration()]
)

Database connections being left open

Python: 3.6.5
Django: 2.1.2
sentry-sdk: 0.4.2

I upgraded from raven to sentry-sdk yesterday and noticed my app was now keeping database connections open rather than closing them per request.

I wrote a tiny view to test this:

def sentry_test(request):
    from django.contrib.auth.models import User
    print(User.objects.all())
    return HttpResponse("<h1>Hello</h1>")

and when GET the URL repeatedly I see the subsequent connections remain idle after the request has completed:

=> SELECT state, query FROM pg_stat_activity WHERE client_addr IS NOT NULL;
 state  | query
--------+-----------------------------------------------------
 idle   | SELECT "auth_user"."id", "auth_user"."password", ...
 active | SELECT state, query FROM pg_stat_activity WHERE client_addr IS NOT NULL;
 idle   | SELECT "auth_user"."id", "auth_user"."password", ...
 idle   | SELECT "django_migrations"."app", "django_migrations"."name" FROM "django_migrations"
 idle   | SELECT "auth_user"."id", "auth_user"."password", ...
 idle   | SELECT "auth_user"."id", "auth_user"."password", ...
 idle   | SELECT "auth_user"."id", "auth_user"."password", ...

As you can see the connection from the migrate query on startup has remained open too.

If I use raven this behaviour is not exhibited.

I suspect this won't help much, but if I use sentry-sdk and comment out:

        WSGIHandler.__call__ = sentry_patched_wsgi_handler

Then the problem doesn't manifest itself but I've basically turned off sentry at this stage.

Notes for Django

• Probably should try to find an earlier entry point other than settings.py (but I'm out of touch w/ Django's load process)
• sentry_sdk.init is better than init on our suggest import process

Test Plan:

• catches wsgi error?
• catches manage.py error?
• catches sys.except_hook (non manage.py) error?
• catches middleware error?
• catches view error?

Traceback line number does not match error report line number

Pull Request #30

Sometimes an error report line is reported incorrectly

tb.tb_fame.f_lineno may differ in some way.
Must use tb.tb_lineno

sentry-sdk/event.py

class Event(Mapping):
    __slots__ = ("_data", "_exc_value")

    def __init__(self, data={}):
        self._data = {
            "event_id": uuid.uuid4().hex,
            "timestamp": datetime.datetime.utcnow(),
            "level": "error",
        }

        self._data.update(data)

        self._exc_value = None

    def set_exception(self, exc_type, exc_value, tb, with_locals):
        print("----------------- trace back --------------------")
        print("f_lineno:  " + str(tb.tb_next.tb_next.tb_frame.f_lineno))
        print("tb_lineno: " + str(tb.tb_next.tb_next.tb_lineno))
        print("f_lineno:  " + str(tb.tb_next.tb_next.tb_next.tb_frame.f_lineno))
        print("tb_lineno: " + str(tb.tb_next.tb_next.tb_next.tb_lineno))
        print("f_lineno:  " + str(tb.tb_next.tb_next.tb_next.tb_next.tb_frame.f_lineno))
        print("tb_lineno: " + str(tb.tb_next.tb_next.tb_next.tb_next.tb_lineno))
        print("f_lineno:  " + str(tb.tb_next.tb_next.tb_next.tb_next.tb_next.tb_frame.f_lineno))
        print("tb_lineno: " + str(tb.tb_next.tb_next.tb_next.tb_next.tb_next.tb_lineno))
        print("f_lineno:  " + str(tb.tb_next.tb_next.tb_next.tb_next.tb_next.tb_next.tb_frame.f_lineno))
        print("tb_lineno: " + str(tb.tb_next.tb_next.tb_next.tb_next.tb_next.tb_next.tb_lineno))
        print("f_lineno:  " + str(tb.tb_next.tb_next.tb_next.tb_next.tb_next.tb_next.tb_next.tb_frame.f_lineno))
        print("tb_lineno: " + str(tb.tb_next.tb_next.tb_next.tb_next.tb_next.tb_next.tb_next.tb_lineno))
        print("f_lineno:  " + str(tb.tb_next.tb_next.tb_next.tb_next.tb_next.tb_next.tb_next.tb_next.tb_frame.f_lineno))
        print("tb_lineno: " + str(tb.tb_next.tb_next.tb_next.tb_next.tb_next.tb_next.tb_next.tb_next.tb_lineno))
        print("----------------- trace back --------------------")
        self["exception"] = {
            "values": exceptions_from_error_tuple(exc_type, exc_value, tb, with_locals)
        }
        self._exc_value = exc_value

trace back

        File "example.py", line 128, in _get_response
          response = self.process_exception_by_middleware(e, request)
        File "example.py", line 126, in _get_response
          response = wrapped_callback(request, *callback_args, **callback_kwargs)
        File "example.py", line 54, in wrapped_view
          return view_func(*args, **kwargs)
        File "example.py", line 69, in view
          return self.dispatch(request, *args, **kwargs)
        File "example.py", line 483, in dispatch
          response = self.handle_exception(exc)
        File "example.py", line 443, in handle_exception
          self.raise_uncaught_exception(exc)
        File "example.py", line 480, in dispatch
          response = handler(request, *args, **kwargs)
        File "example.py", line 210, in post

output

----------------- trace back --------------------
f_lineno:  128
tb_lineno: 126
f_lineno:  54
tb_lineno: 54
f_lineno:  69
tb_lineno: 69
f_lineno:  483
tb_lineno: 483
f_lineno:  443
tb_lineno: 443
f_lineno:  483
tb_lineno: 480
f_lineno:  213
tb_lineno: 210
----------------- trace back --------------------

Add extra data to event or message

With raven-python we could add extra param to captureMessage. Now with sentry-python capture_message does not accept extra param.

Can we add extra data to event or message ?

Ensure the Django shell does not send its errors to sentry

The logging integration will capture the errors from the REPL and send them to sentry. We should try to detect those loggers and ignore them entirely.

Celery - double errors logged when task decorated

I have some tasks that require distributed lock mechanism and it is implemented as decorator for celery tasks:

@app.task
@celery_lock('receive_tasks', ttl=40)
def receive_tasks(ad_id):
    raise Exception('sentry')

When exception is raised i get two error events logged, one is with tags: handled=no,level=error,mechanism=celery,server_name=,transaction=.receive_tasks
and warning text:

There was 1 error encountered while processing this event
Discarded invalid parameter '' Expand

Second event is almost equal to first one, but with tags: handled=yes, level=error, logger=celery.worker.job, mechanism=logging, server_name=
And contains message from celery worker stdout:

Task .receive_tasks[f209f4c9-61cb-47b8-94ea-b15e69933897] raised unexpected: Exception('sentry',)

If i remove @celery_lock decorator from task or pass lock_id as named argument @celery_lock(lock_id='receive_tasks', ttl=40), then all works fine - just one event received and no warnings.

Seems that this problem shows up because of non-keyword decorator args.

celery_lock implemented as simple decorator:

def celery_lock(lock_id, ttl):
    def decorator(func):
        @wraps(func)
        def lock_decorator(*args, **kwargs):
            with CeleryLock(lock_id, ttl) as aq:
                if not aq:
                    return 'locked'
                return func(*args, **kwargs)
        return lock_decorator
    return decorator

AWS lambda integration

attach_stacktrace is broken in master

I misunderstood what the option meant. Do not release for now. gtg

Looks like the raven module still needs to be installed ?

Hello !

I just switched to using sentry-sdk instead of raven.

However I was getting the following error:

2018-10-17 13:14:54,907: Error running WSGI application
2018-10-17 13:14:54,913: ValueError: Unable to configure handler 'sentry': Cannot resolve 'raven.contrib.django.raven_compat.handlers.SentryHandler': No module named 'raven'
2018-10-17 13:14:54,915:   File "/var/www/www_mysite_com_wsgi.py", line 18, in <module>
2018-10-17 13:14:54,917:     application = get_wsgi_application()

Which was fixed by reinstalling raven via pip.

Is it intended for raven to be a requirement ? It's not mentioned in either the Python or Django related pages: https://docs.sentry.io/platforms/python/django/

Aiohttp support

I don't see a clear way to integrate with aiohttp (server) just yet. I can still use raven at present, but I imagine sentry-python/sentry-sdk is the proposed way forward longer term?

Put sys.argv into extra again

Raven used to put sys.argv as extra key. This is useful for CLI apps

Logging customizations

In the previous version of the sdk, with django I've specialized logging handlers for sentry, this allows for example, that some specific logging always target sentry.

Now I see that only the level control if the message ends in sentry.

Is there any plan of adding the previous feature to the new client?

Django Template Tag {% sentry_public_dsn %} missing

While moving a project from raven-python to sentry-python I noticed, that there is no documentation (and maybe implementation?) for the template tag {% sentry_public_dsn %}.
It was available in raven-python - I found it to be quiet handy.
I hope I am not missing something - or is there a reason that this wasn't implemented?

Rq support

We should bring back support for rq.

Support threadless Python environments

Is there any support to transport synchronously like HTTPTransport of raven (https://docs.sentry.io/clients/python/transports/#sync)? I could not find any support as far as I read code.
I want to transport synchronously for AWS Batch. AWS Batch can not send any exceptions in asynchronous way because it stop as soon as it raises exceptions.

Wrong version number reported

I have sentry-sdk==0.3.6 installed, but events in the UI show up as

Add `capture_exceptions` decorator

Please, I'm missing this feature from raven.
https://github.com/getsentry/raven-python/blob/4206873d63c461ffe6944029d36951514157928e/raven/base.py#L842

Rename meta key to _meta

sentry.io cannot handle "" as meta key, only "_meta"

Match old Transactions in Django

We only report the local function name in old transactions for django. David wants this to be the URL rule for Django. We should port the old resolver over to it (https://github.com/getsentry/raven-python/blob/03559bb05fd963e2be96372ae89fb0bce751d26d/raven/contrib/django/resolver.py).

Catch crashing threads

since sys.excepthook is not called for threads that crash, we should write a separate integration for this (turned on by default)

How to extract more info from django request?

The user data that's logged by default (with sending of PII enabled) looks something like

{'username': teeberg', 'ip_address': '127.0.0.1', 'email': [email protected]'}

I'm trying to add some fields to this, but there doesn't seem to be an official documented way to do this.

Is there a cleaner way than monkey-patching sentry_sdk.integrations.django._set_user_info?

Maybe the request could be sent along as another hint to before_send handlers?

Missing integrations

• httplib for breadcrumbs
• requests for breadcrumbs
• aws lambda
• sanic

Invalid project in DSN

I have a sentry server set up on premises. I did this a couple weeks ago, just before you guys updated the docs with the new unified SDK.

Now I'm updating our project (Django) to use the python sdk. The problem is that in our project in sentry, the DSN I see has this format:

http://[email protected]:9000//2

Now the problem comes when initializing the integration. This DSN format (note: the double slash there) makes the initialization fail in sentry_sdk/utils.py line 77, when int('/2') is attempted.

This is easily fixed by simply removing the extra slash, but might be a headache for anyone who doesn't notice that.

Not sure if the SDK should handle this or if this is an issue with Sentry itself, but at least this didn't use to happen with Raven.

Sanic integration

Possible double reporting in 0.3.7

We have a simple test that verifies the Sentry SDK is capturing Flask errors because we had some configuration/PyPy issues in the past:

https://github.com/getsentry/snuba/blob/fd44ec8880240575ee2cb05bfc87e4cfd1d0e853/tests/test_api.py#L928-L936

As of this writing we don't pin the sentry-sdk version:

https://github.com/getsentry/snuba/blob/fd44ec8880240575ee2cb05bfc87e4cfd1d0e853/setup.py#L80

The test started failing because it seems that as of 0.3.7 (I tried 0.3.6 and wasn't able to reproduce the issue) the transport is being called twice for the same exception, causing the number of reported issues to be 2 instead of 1.

Maybe this is a false alarm but I think our integration is relatively simple so this may be signs of a larger issue?

Add debug mode and do not use _internal_exceptions on expected fails

A debug mode should be added that logs out when events are dropped and internal exceptions are ignored.

CeleryIntegration is not capturing locals

sentry-sdk==0.3.9
When Exception happens inside celery task, I can see traceback in sentry, but without local variables.
As a workaround, I am capturing exception directly, in BaseTask and it works.

class BaseTask(Task):
    def on_failure(self, exc, task_id, *args, **kwargs):
        """When passing raw exception to sentry we will get local variable values instead of just message"""
        with push_scope() as scope:
            scope.set_extra('task_id', task_id)
            scope.set_extra('retries', self.request.retries)
            capture_exception(exc)
        super().on_failure(exc, task_id, *args, **kwargs)

usage: @shared_task(base=BaseTask)

In CeleryIntegration def _process_failure_signal(self, sender, task_id, einfo, **kw): gets einfo, so I don't know why it's not working out of the box.

AWS Chalice support

Chalice defines itself as "Python Serverless Microframework for AWS": https://github.com/aws/chalice

Some relevant information here: #110.

---

EDIT: Please vote on this issue with 👍 for Chalice support.

Getting Attribute Error 'list' object has no attribute 'popleft' in hub.py add_breadcrumb

I have followed installation instructions mentioned in this page
https://docs.sentry.io/quickstart?platform=python

After a while, sentry fails and gives this error message.

File "/app/.heroku/python/lib/python3.6/site-packages/sentry_sdk/api.py", line 91, in add_breadcrumb
   return hub.add_breadcrumb(*args, **kwargs)
File "/app/.heroku/python/lib/python3.6/site-packages/sentry_sdk/hub.py", line 206, in add_breadcrumb
   scope._breadcrumbs.popleft()
AttributeError: 'list' object has no attribute 'popleft'

https://github.com/getsentry/sentry-python/blob/master/sentry_sdk/hub.py#L209

ignore_loggers should be per-client

A client should not ignore the django logger if the client does not have the django integration enabled.

ModuleNotFoundError: No module named 'runtime'

When trying to import AwsLambdaIntegration I get following Error:

ModuleNotFoundError: No module named 'runtime'

Traceback:

ModuleNotFoundError                       Traceback (most recent call last)
<ipython-input-1-f0a35f42fdce> in <module>()
----> 1 from sentry_sdk.integrations.aws_lambda import AwsLambdaIntegration

~/Envs/gds/lib/python3.6/site-packages/sentry_sdk/integrations/aws_lambda.py in <module>()
     13 
     14 import __main__ as lambda_bootstrap
---> 15 import runtime as lambda_runtime
     16 
     17 

ModuleNotFoundError: No module named 'runtime'

Version:

sentry-sdk==0.3.9

I tried install runtime package, but it didn't help, because then I get:

~/Envs/gds/lib/python3.6/site-packages/runtime.py in <module>()
     26         from thread import get_ident
     27     except ImportError:  # pragma: no cover
---> 28         from dummy_thread import get_ident
     29 
     30 __all__ = (

ModuleNotFoundError: No module named 'dummy_thread'

And pip install thread returns:

No matching distribution found for thread

Patch Django to catch manage.py errors

We should be able to capture errors resulting from a crashing manage.py task, see #19

Please, fix AttributeError: 'list' object has no attribute 'popleft'

Python 3.7.0, sentry_sdk 0.3.2

2018-09-21 18:08:21,575 DEBUG Internal error in sentry_sdk
Traceback (most recent call last):
  File "/home/dmugtasimov/.virtualenvs/packy-agent2-eBU3dgqa/lib/python3.7/site-packages/sentry_sdk/utils.py", line 28, in capture_internal_exceptions
    yield
  File "/home/dmugtasimov/.virtualenvs/packy-agent2-eBU3dgqa/lib/python3.7/site-packages/sentry_sdk/integrations/logging.py", line 102, in _emit
    self._breadcrumb_from_record(record), hint={"log_record": record}
  File "/home/dmugtasimov/.virtualenvs/packy-agent2-eBU3dgqa/lib/python3.7/site-packages/sentry_sdk/hub.py", line 209, in add_breadcrumb
    scope._breadcrumbs.popleft()
AttributeError: 'list' object has no attribute 'popleft'

asyncio concurrency issues

@mitsuhiko discovered that passing around hubs does not work at all and we likely leak state between requests