Skip to main content

Overview

Serialization library for Exceptions and Tracebacks.

  • Free software: BSD license

It allows you to:

  • Pickle tracebacks and raise exceptions with pickled tracebacks in different processes. This allows better error handling when running code over multiple processes (imagine multiprocessing, billiard, futures, celery etc).

  • Create traceback objects from strings (the from_string method). No pickling is used.

  • Serialize tracebacks to/from plain dicts (the from_dict and to_dict methods). No pickling is used.

  • Raise the tracebacks created from the aforementioned sources.

  • Pickle an Exception together with its traceback and exception chain (raise ... from ...) (Python 3 only)

Again, note that using the pickle support is completely optional. You are solely responsible for security problems should you decide to use the pickle support.

Installation

pip install tblib

Documentation

Pickling tracebacks

Note: The traceback objects that come out are stripped of some attributes (like variables). But you’ll be able to raise exceptions with those tracebacks or print them - that should cover 99% of the usecases.

>>> from tblib import pickling_support
>>> pickling_support.install()
>>> import pickle, sys
>>> def inner_0():
...     raise Exception('fail')
...
>>> def inner_1():
...     inner_0()
...
>>> def inner_2():
...     inner_1()
...
>>> try:
...     inner_2()
... except:
...     s1 = pickle.dumps(sys.exc_info())
...
>>> len(s1) > 1
True
>>> try:
...     inner_2()
... except:
...     s2 = pickle.dumps(sys.exc_info(), protocol=pickle.HIGHEST_PROTOCOL)
...
>>> len(s2) > 1
True

>>> try:
...     import cPickle
... except ImportError:
...     import pickle as cPickle
>>> try:
...     inner_2()
... except:
...     s3 = cPickle.dumps(sys.exc_info(), protocol=pickle.HIGHEST_PROTOCOL)
...
>>> len(s3) > 1
True

Unpickling tracebacks

>>> pickle.loads(s1)
(<...Exception'>, Exception('fail'...), <traceback object at ...>)

>>> pickle.loads(s2)
(<...Exception'>, Exception('fail'...), <traceback object at ...>)

>>> pickle.loads(s3)
(<...Exception'>, Exception('fail'...), <traceback object at ...>)

Raising

>>> from tblib.decorators import reraise
>>> reraise(*pickle.loads(s1))
Traceback (most recent call last):
  ...
  File "<doctest README.rst[14]>", line 1, in <module>
    reraise(*pickle.loads(s2))
  File "<doctest README.rst[8]>", line 2, in <module>
    inner_2()
  File "<doctest README.rst[5]>", line 2, in inner_2
    inner_1()
  File "<doctest README.rst[4]>", line 2, in inner_1
    inner_0()
  File "<doctest README.rst[3]>", line 2, in inner_0
    raise Exception('fail')
Exception: fail
>>> reraise(*pickle.loads(s2))
Traceback (most recent call last):
  ...
  File "<doctest README.rst[14]>", line 1, in <module>
    reraise(*pickle.loads(s2))
  File "<doctest README.rst[8]>", line 2, in <module>
    inner_2()
  File "<doctest README.rst[5]>", line 2, in inner_2
    inner_1()
  File "<doctest README.rst[4]>", line 2, in inner_1
    inner_0()
  File "<doctest README.rst[3]>", line 2, in inner_0
    raise Exception('fail')
Exception: fail
>>> reraise(*pickle.loads(s3))
Traceback (most recent call last):
  ...
  File "<doctest README.rst[14]>", line 1, in <module>
    reraise(*pickle.loads(s2))
  File "<doctest README.rst[8]>", line 2, in <module>
    inner_2()
  File "<doctest README.rst[5]>", line 2, in inner_2
    inner_1()
  File "<doctest README.rst[4]>", line 2, in inner_1
    inner_0()
  File "<doctest README.rst[3]>", line 2, in inner_0
    raise Exception('fail')
Exception: fail

Pickling Exceptions together with their traceback and chain (Python 3 only)

>>> try:  # doctest: +SKIP
...     try:
...         1 / 0
...     except Exception as e:
...         raise Exception("foo") from e
... except Exception as e:
...     s = pickle.dumps(e)
>>> raise pickle.loads(s)  # doctest: +SKIP
Traceback (most recent call last):
  File "<doctest README.rst[16]>", line 3, in <module>
    1 / 0
ZeroDivisionError: division by zero

The above exception was the direct cause of the following exception:

Traceback (most recent call last):
  File "<doctest README.rst[17]>", line 1, in <module>
    raise pickle.loads(s)
  File "<doctest README.rst[16]>", line 5, in <module>
    raise Exception("foo") from e
Exception: foo

BaseException subclasses defined after calling pickling_support.install() will not retain their traceback and exception chain pickling. To cover custom Exceptions, there are three options:

  1. Use @pickling_support.install as a decorator for each custom Exception

    >>> from tblib import pickling_support
    >>> # Declare all imports of your package's dependencies
    >>> import numpy  # doctest: +SKIP
    
    >>> pickling_support.install()  # install for all modules imported so far
    
    >>> @pickling_support.install
    ... class CustomError(Exception):
    ...     pass

    Eventual subclasses of CustomError will need to be decorated again.

  2. Invoke pickling_support.install() after all modules have been imported and all Exception subclasses have been declared

    >>> # Declare all imports of your package's dependencies
    >>> import numpy  # doctest: +SKIP
    >>> from tblib import pickling_support
    
    >>> # Declare your own custom Exceptions
    >>> class CustomError(Exception):
    ...     pass
    
    >>> # Finally, install tblib
    >>> pickling_support.install()
  3. Selectively install tblib for Exception instances just before they are pickled

    pickling_support.install(<Exception instance>, [Exception instance], ...)

    The above will install tblib pickling for all listed exceptions as well as any other exceptions in their exception chains.

    For example, one could write a wrapper to be used with ProcessPoolExecutor, Dask.distributed, or similar libraries:

>>> from tblib import pickling_support
>>> def wrapper(func, *args, **kwargs):
...     try:
...         return func(*args, **kwargs)
...     except Exception as e:
...         pickling_support.install(e)
...         raise

What if the paths on the other side are different?

Tracebacks pickled on another machine, in a container or in a sandbox refer to files by the paths they had there. When rendering them locally you can map those paths onto local ones with pickling_support.map_filenames (or Traceback.as_traceback(map_filename=...) if you are using tblib.Traceback directly):

>>> with pickling_support.map_filenames(lambda name: name.replace('<doctest', '<remote doctest')):
...     exc_info = pickle.loads(s3)
>>> exc_info[2].tb_frame.f_code.co_filename
'<remote doctest README.rst[11]>'

What if we have a local stack, does it show correctly ?

Yes it does:

>>> exc_info = pickle.loads(s3)
>>> def local_0():
...     reraise(*exc_info)
...
>>> def local_1():
...     local_0()
...
>>> def local_2():
...     local_1()
...
>>> local_2()
Traceback (most recent call last):
  File "...doctest.py", line ..., in __run
    compileflags, 1) in test.globs
  File "<doctest README.rst[24]>", line 1, in <module>
    local_2()
  File "<doctest README.rst[23]>", line 2, in local_2
    local_1()
  File "<doctest README.rst[22]>", line 2, in local_1
    local_0()
  File "<doctest README.rst[21]>", line 2, in local_0
    reraise(*exc_info)
  File "<doctest README.rst[11]>", line 2, in <module>
    inner_2()
  File "<doctest README.rst[5]>", line 2, in inner_2
    inner_1()
  File "<doctest README.rst[4]>", line 2, in inner_1
    inner_0()
  File "<doctest README.rst[3]>", line 2, in inner_0
    raise Exception('fail')
Exception: fail

It also supports more contrived scenarios

Like tracebacks with syntax errors:

>>> from tblib import Traceback
>>> from examples import bad_syntax
>>> try:
...     bad_syntax()
... except:
...     et, ev, tb = sys.exc_info()
...     tb = Traceback(tb)
...
>>> reraise(et, ev, tb.as_traceback())
Traceback (most recent call last):
  ...
  File "<doctest README.rst[58]>", line 1, in <module>
    reraise(et, ev, tb.as_traceback())
  File "<doctest README.rst[57]>", line 2, in <module>
    bad_syntax()
  File "...tests...examples.py", line 18, in bad_syntax
    import badsyntax
  File "...tests...badsyntax.py", line 5
    is very bad
     ^
SyntaxError: invalid syntax

Or other import failures:

>>> from examples import bad_module
>>> try:
...     bad_module()
... except:
...     et, ev, tb = sys.exc_info()
...     tb = Traceback(tb)
...
>>> reraise(et, ev, tb.as_traceback())
Traceback (most recent call last):
  ...
  File "<doctest README.rst[61]>", line 1, in <module>
    reraise(et, ev, tb.as_traceback())
  File "<doctest README.rst[60]>", line 2, in <module>
    bad_module()
  File "...tests...examples.py", line 23, in bad_module
    import badmodule
  File "...tests...badmodule.py", line 3, in <module>
    raise Exception("boom!")
Exception: boom!

Or a traceback that’s caused by exceeding the recursion limit (here we’re forcing the type and value to have consistency across platforms):

>>> def f(): f()
>>> try:
...    f()
... except RuntimeError:
...    et, ev, tb = sys.exc_info()
...    tb = Traceback(tb)
...
>>> reraise(RuntimeError, RuntimeError("maximum recursion depth exceeded"), tb.as_traceback())
Traceback (most recent call last):
  ...
  File "<doctest README.rst[32]>", line 1, in f
    def f(): f()
  File "<doctest README.rst[32]>", line 1, in f
    def f(): f()
  File "<doctest README.rst[32]>", line 1, in f
    def f(): f()
  ...
RuntimeError: maximum recursion depth exceeded

Reference

tblib.Traceback

It is used by the pickling_support. You can use it too if you want more flexibility:

>>> from tblib import Traceback
>>> try:
...     inner_2()
... except:
...     et, ev, tb = sys.exc_info()
...     tb = Traceback(tb)
...
>>> reraise(et, ev, tb.as_traceback())
Traceback (most recent call last):
  ...
  File "<doctest README.rst[21]>", line 6, in <module>
    reraise(et, ev, tb.as_traceback())
  File "<doctest README.rst[21]>", line 2, in <module>
    inner_2()
  File "<doctest README.rst[5]>", line 2, in inner_2
    inner_1()
  File "<doctest README.rst[4]>", line 2, in inner_1
    inner_0()
  File "<doctest README.rst[3]>", line 2, in inner_0
    raise Exception('fail')
Exception: fail
tblib.Traceback.to_dict

You can use the to_dict method and the from_dict classmethod to convert a Traceback into and from a dictionary serializable by the stdlib json.JSONDecoder:

>>> import json
>>> try:
...     inner_2()
... except:
...     et, ev, tb = sys.exc_info()
...     tb = Traceback(tb)
...     tb_dict = tb.to_dict()
...     'tb_frame' in tb_dict and 'tb_lineno' in tb_dict
True
tblib.Traceback.from_dict

Building on the previous example:

>>> tb_json = json.dumps(tb_dict)
>>> tb = Traceback.from_dict(json.loads(tb_json))
>>> reraise(et, ev, tb.as_traceback())
Traceback (most recent call last):
  ...
  File "<doctest README.rst[21]>", line 6, in <module>
    reraise(et, ev, tb.as_traceback())
  File "<doctest README.rst[21]>", line 2, in <module>
    inner_2()
  File "<doctest README.rst[5]>", line 2, in inner_2
    inner_1()
  File "<doctest README.rst[4]>", line 2, in inner_1
    inner_0()
  File "<doctest README.rst[3]>", line 2, in inner_0
    raise Exception('fail')
Exception: fail
tblib.Traceback.from_string
>>> tb = Traceback.from_string("""
... File "skipped.py", line 123, in func_123
... Traceback (most recent call last):
...   File "tests/examples.py", line 2, in func_a
...     func_b()
...   File "tests/examples.py", line 6, in func_b
...     func_c()
...   File "tests/examples.py", line 10, in func_c
...     func_d()
...   File "tests/examples.py", line 14, in func_d
... Doesn't: matter
... """)
>>> reraise(et, ev, tb.as_traceback())
Traceback (most recent call last):
  ...
  File "<doctest README.rst[42]>", line 6, in <module>
    reraise(et, ev, tb.as_traceback())
  File "...examples.py", line 2, in func_a
    func_b()
  File "...examples.py", line 6, in func_b
    func_c()
  File "...examples.py", line 10, in func_c
    func_d()
  File "...examples.py", line 14, in func_d
    raise Exception('Guessing time !')
Exception: fail

If you use the strict=False option then parsing is a bit more lax:

>>> tb = Traceback.from_string("""
... File "bogus.py", line 123, in bogus
... Traceback (most recent call last):
...  File "tests/examples.py", line 2, in func_a
...   func_b()
...    File "tests/examples.py", line 6, in func_b
...     func_c()
...    File "tests/examples.py", line 10, in func_c
...   func_d()
...  File "tests/examples.py", line 14, in func_d
... Doesn't: matter
... """, strict=False)
>>> reraise(et, ev, tb.as_traceback())
Traceback (most recent call last):
  ...
  File "<doctest README.rst[42]>", line 6, in <module>
    reraise(et, ev, tb.as_traceback())
  File "bogus.py", line 123, in bogus
  File "...examples.py", line 2, in func_a
    func_b()
  File "...examples.py", line 6, in func_b
    func_c()
  File "...examples.py", line 10, in func_c
    func_d()
  File "...examples.py", line 14, in func_d
    raise Exception('Guessing time !')
Exception: fail

tblib.decorators.return_error

>>> from tblib.decorators import return_error
>>> inner_2r = return_error(inner_2)
>>> e = inner_2r()
>>> e
<tblib.decorators.Error object at ...>
>>> e.reraise()
Traceback (most recent call last):
  ...
  File "<doctest README.rst[26]>", line 1, in <module>
    e.reraise()
  File "...tblib...decorators.py", line 19, in reraise
    reraise(self.exc_type, self.exc_value, self.traceback)
  File "...tblib...decorators.py", line 25, in return_exceptions_wrapper
    return func(*args, **kwargs)
  File "<doctest README.rst[5]>", line 2, in inner_2
    inner_1()
  File "<doctest README.rst[4]>", line 2, in inner_1
    inner_0()
  File "<doctest README.rst[3]>", line 2, in inner_0
    raise Exception('fail')
Exception: fail

How’s this useful? Imagine you’re using multiprocessing like this:

# Note that Python 3.4 and later will show the remote traceback (but as a string sadly) so we skip testing this.
>>> import traceback
>>> from multiprocessing import Pool
>>> from examples import func_a
>>> pool = Pool()  # doctest: +SKIP
>>> try:  # doctest: +SKIP
...     for i in pool.map(func_a, range(5)):
...         print(i)
... except:
...     print(traceback.format_exc())
...
Traceback (most recent call last):
  File "<doctest README.rst[...]>", line 2, in <module>
    for i in pool.map(func_a, range(5)):
  File "...multiprocessing...pool.py", line ..., in map
    ...
  File "...multiprocessing...pool.py", line ..., in get
    ...
Exception: Guessing time !
<BLANKLINE>
>>> pool.terminate()  # doctest: +SKIP

Not very useful is it? Let’s sort this out:

>>> from tblib.decorators import apply_with_return_error, Error
>>> from itertools import repeat
>>> pool = Pool()
>>> try: # doctest: +SKIP
...     for i in pool.map(apply_with_return_error, zip(repeat(func_a), range(5))):
...         if isinstance(i, Error):
...             i.reraise()
...         else:
...             print(i)
... except:
...     print(traceback.format_exc())
...
Traceback (most recent call last):
  File "<doctest README.rst[...]>", line 4, in <module>
    i.reraise()
  File "...tblib...decorators.py", line ..., in reraise
    reraise(self.exc_type, self.exc_value, self.traceback)
  File "...tblib...decorators.py", line ..., in reraise
    raise value.with_traceback(tb)
  File "...tblib...decorators.py", line ..., in return_exceptions_wrapper
    return func(*args, **kwargs)
  File "...tblib...decorators.py", line ..., in apply_with_return_error
    return args[0](*args[1:])
  File "...examples.py", line 2, in func_a
    func_b()
  File "...examples.py", line 6, in func_b
    func_c()
  File "...examples.py", line 10, in func_c
    func_d()
  File "...examples.py", line 14, in func_d
    raise Exception('Guessing time !')
Exception: Guessing time !
<BLANKLINE>
>>> pool.terminate()

Much better !

What if we have a local call stack ?
>>> def local_0():
...     pool = Pool()
...     try:
...         for i in pool.map(apply_with_return_error, zip(repeat(func_a), range(5))):
...             if isinstance(i, Error):
...                 i.reraise()
...             else:
...                 print(i)
...     finally:
...         pool.close()
...
>>> def local_1():
...     local_0()
...
>>> def local_2():
...     local_1()
...
>>> try: # doctest: +SKIP
...     local_2()
... except:
...     print(traceback.format_exc())
...
Traceback (most recent call last):
  File "<doctest README.rst[...]>", line 2, in <module>
    local_2()
  File "<doctest README.rst[...]>", line 2, in local_2
    local_1()
  File "<doctest README.rst[...]>", line 2, in local_1
    local_0()
  File "<doctest README.rst[...]>", line 6, in local_0
    i.reraise()
  File "...tblib...decorators.py", line ..., in reraise
    reraise(self.exc_type, self.exc_value, self.traceback)
  File "...tblib...decorators.py", line ..., in reraise
    raise value.with_traceback(tb)
  File "...tblib...decorators.py", line ..., in return_exceptions_wrapper
    return func(*args, **kwargs)
  File "...tblib...decorators.py", line ..., in apply_with_return_error
    return args[0](*args[1:])
  File "...tests...examples.py", line 2, in func_a
    func_b()
  File "...tests...examples.py", line 6, in func_b
    func_c()
  File "...tests...examples.py", line 10, in func_c
    func_d()
  File "...tests...examples.py", line 14, in func_d
    raise Exception('Guessing time !')
Exception: Guessing time !
<BLANKLINE>
Other weird stuff

Clearing traceback works (Python 3.4 and up):

>>> tb = Traceback.from_string("""
... File "skipped.py", line 123, in func_123
... Traceback (most recent call last):
...   File "tests/examples.py", line 2, in func_a
...     func_b()
...   File "tests/examples.py", line 6, in func_b
...     func_c()
...   File "tests/examples.py", line 10, in func_c
...     func_d()
...   File "tests/examples.py", line 14, in func_d
... Doesn't: matter
... """)
>>> import traceback, sys
>>> if sys.version_info > (3, 4):
...     traceback.clear_frames(tb)

Credits

Changelog

3.2.3 (2026-09-18)

  • Reconstructed tracebacks now carry the position information of the original expression on Python 3.11+, so traceback and pytest print the same caret markers as for the original exception, including the ~~^~~ anchors and multi-line expressions. Fixes #76.

  • Code objects record co_firstlineno and as_dict/from_dict carry it, so rebuilt frames report the function’s real first line instead of the failing line. This lets pytest show the function body up to the failure as it does for native tracebacks. Dictionaries without the key still load.

  • Tracebacks reconstructed from dicts or strings without position information no longer gain bogus positions from the stub (tb_lasti is -1 in that case).

  • Added Traceback.as_traceback(map_filename=...) and the pickling_support.map_filenames() context manager to rewrite frame file names while a traceback is rebuilt, for tracebacks produced under different paths.

  • Reconstruction no longer depends on the contents of the original f_globals/f_locals and works on PyPy without implementation-specific co_linetable encoding.

3.2.2 (2025-11-12)

  • Fixed regression occurring with TimeoutError exceptions. They should be represented now exactly as the original when unpickling. Contributed by Jacob Tomlinson in #85.

3.2.1 (2025-10-31)

  • Fixed regression occurring with ExceptionGroup exceptions. That exception type is now handled specifically in the new unpickle_exception_with_attrs function (just like OSError).

3.2.0 (2025-10-21)

  • Changed tblib.pickling_support.install to support exceptions with __init__ that does match the default BaseException.__reduce__ (as it expects the positional arguments to __init__ to match the args attribute).

    Special handling for OSError (and subclasses) is also included. The errno, strerror, winerror, filename and filename2 attributes will be added in the reduce structure (if set).

    This will support exception subclasses that do this without defining a custom __reduce__:

    def __init__(self):
        super().__init__('mistery argument')
    
    def __init__(self, mistery_argument):
        super().__init__()
        self.mistery_argument = mistery_argument

    Tests and POC contributed by Oldřich Jedlička in #73.

  • Fixed some doctest and coverage config. Contributed by Colin Watson in #79.

3.1.0 (2025-03-31)

  • Improved performance of as_traceback by a large factor. Contributed by Haoyu Weng in #81.

  • Dropped support for now-EOL Python 3.8 and added 3.13 in the test grid.

3.0.0 (2023-10-22)

  • Added support for __context__, __suppress_context__ and __notes__. Contributed by Tim Maxwell in #72.

  • Added the get_locals argument to tblib.pickling_support.install(), tblib.Traceback and tblib.Frame. Fixes #41.

  • Dropped support for now-EOL Python 3.7 and added 3.12 in the test grid.

2.0.0 (2023-06-22)

  • Removed support for legacy Pythons (2.7 and 3.6) and added Python 3.11 in the test grid.

  • Some cleanups and refactors (mostly from ruff).

1.7.0 (2020-07-24)

  • Add more attributes to Frame and Code objects for pytest compatibility. Contributed by Ivanq in #58.

1.6.0 (2019-12-07)

  • When pickling an Exception, also pickle its traceback and the Exception chain (raise ... from ...). Contributed by Guido Imperiale in #53.

1.5.0 (2019-10-23)

  • Added support for Python 3.8. Contributed by Victor Stinner in #42.

  • Removed support for end of life Python 3.4.

  • Few CI improvements and fixes.

1.4.0 (2019-05-02)

  • Removed support for end of life Python 3.3.

  • Fixed tests for Python 3.7. Contributed by Elliott Sales de Andrade in #36.

  • Fixed compatibility issue with Twised (twisted.python.failure.Failure expected a co_code attribute).

1.3.2 (2017-04-09)

  • Add support for PyPy3.5-5.7.1-beta. Previously AttributeError: 'Frame' object has no attribute 'clear' could be raised. See PyPy issue #2532.

1.3.1 (2017-03-27)

  • Fixed handling for tracebacks due to exceeding the recursion limit. Fixes #15.

1.3.0 (2016-03-08)

  • Added Traceback.from_string.

1.2.0 (2015-12-18)

  • Fixed handling for tracebacks from generators and other internal improvements and optimizations. Contributed by DRayX in #10 and #11.

1.1.0 (2015-07-27)

  • Added support for Python 2.6. Contributed by Arcadiy Ivanov in #8.

1.0.0 (2015-03-30)

  • Added to_dict method and from_dict classmethod on Tracebacks. Contributed by beckjake in #5.

Metadata

Release files for pyodide-tblib 3.2.3

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for pyodide-tblib 3.2.3
File Size Uploaded
pyodide_tblib-3.2.3.tar.gz 48.0 kB Details

Built distribution (wheel)

Table of built distributions (wheels) for pyodide-tblib 3.2.3
File Interpreter ABI Platform
pyodide_tblib-3.2.3-py3-none-any.whl Python 3 none any Details

Total release size: 63.9 kB

Release files / pyodide_tblib-3.2.3.tar.gz

Download URL pyodide_tblib-3.2.3.tar.gz
Size 48.0 kB
Tags Source
SHA-256 checksum
How to use checksums
e731c3a8e5fe800ebb990eaa16a48237e6894d46a2291fafdcb854a291e2f6d0
BLAKE2b-256 checksum
How to use checksums
4e8ae30340e31040c9349dfbe059123e7a0276595cacb854f0b613183b27aa4d
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/7.0.0 CPython/3.10.20

Release files / pyodide_tblib-3.2.3-py3-none-any.whl

Download URL pyodide_tblib-3.2.3-py3-none-any.whl
Size 15.9 kB
Tags Python 3
SHA-256 checksum
How to use checksums
2960f305fe3ef1f5fed9ff098c959104af144627a53033a469183200e50d7ea0
BLAKE2b-256 checksum
How to use checksums
6512d91be04a60cd7349cc86358e0915dcfe5edbfcf83875c458e220756872b5
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/7.0.0 CPython/3.10.20

Release history Release notifications | RSS feed

This release

3.2.3 This release

2 release files

1.7.1

2 release files

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page