Celery integration guide

Add BugSnag to your Celery tasks.

Installation

Using PyPI:

pip install bugsnag

The latest available version of bugsnag is v4.7.1.

Basic configuration

  1. Configure the library in your worker module

    import bugsnag
    bugsnag.configure(
        api_key = "YOUR_API_KEY_HERE",
        project_root = "/path/to/your/app"
    )
    

    You can find your API key in Project Settings from your BugSnag dashboard.

  2. Add the bugsnag failure handler to celery

    from bugsnag.celery import connect_failure_handler
    connect_failure_handler()
    

Logging configuration

You can connect BugSnag up to Python’s logging framework so that anything of level “error” or above is logged to BugSnag.

Here is a plain Python example:

import logging

from bugsnag.handlers import BugsnagHandler

# ... (call bugsnag.configure() here)
logger = logging.getLogger("test.logger")
handler = BugsnagHandler()
# send only ERROR-level logs and above
handler.setLevel(logging.ERROR)
logger.addHandler(handler)

To record breadcrumbs for log messages, you can add the leave_breadcrumbs log filter:

from bugsnag.handlers import BugsnagHandler

logger = logging.getLogger("test.logger")
handler = BugsnagHandler()

logger.addFilter(handler.leave_breadcrumbs)

This will capture a breadcrumb for each log message at or above the configured breadcrumb_log_level, which defaults to INFO.

See grouping_hash to understand how you can control the grouping of your log messages.

Reporting unhandled errors

At this point, BugSnag should be installed and configured, and any unhandled exceptions will be automatically detected and should appear in your BugSnag dashboard.

Automatic detection of unhandled exceptions in threads relies on threading.excepthook, which is only supported in Python 3.8 and above.

Reporting handled errors

If you would like to send handled exceptions to BugSnag, you should import the bugsnag module:

import bugsnag

Then to notify BugSnag of an error, you can call bugsnag.notify:

bugsnag.notify(Exception("Something broke!"))

You can also pass additional configuration options in as named parameters. These parameters will only affect the current call to notify.

Avoiding re-notifying exceptions

Sometimes after catching and notifying a handled exception you may want to re-raise the exception to be dealt with by your standard error handlers without sending an automatic exception to BugSnag.

This can be accomplished by calling bugsnag.notify(), adding a custom skip_bugsnag attr to your exception, and then re-raising the exception:

try:
    raise Exception('Something went wrong!')
except Exception as exception:
    bugsnag.notify(exception)
    exception.skip_bugsnag = True

    # Now this won't be sent a second time by the exception handlers
    raise exception

Sending diagnostic data

The metadata field is a dictionary of dictionaries which will be rendered as a tab in a BugSnag error report. This example would create a special_info tab:

bugsnag.notify(Exception("Something broke!"),
    context="myContext",
    metadata={"special_info":{"request_id": 12345, "message_id": 854}}
)

For more information, see reporting handled errors.

Logging diagnostic data

The BugsnagHandler accepts a special keyword argument to its __init__() function: extra_fields. This is an optional dictionary of extra attributes to gather from each LogRecord and insert into metadata to be attached to BugSnag error reports.

The keys in the extra_fields dictionary should be tab names for where you would like the data displayed in BugSnag. The values should be attributes to extract from each log record. The attributes are not required to exist on the log record, and any non-existent attribute will be ignored. For example:

logger = logging.getLogger("your_logger_name")
handler = BugsnagHandler(extra_fields={
    "tab_name": [
        "keyA",
        "keyB"
    ]
})
logger.addHandler(handler)

logger.warning('A warning', extra={
    'keyA': 'abc', # will be added to tab "tab_name"
    'keyB': 'def', # will be added to tab "tab_name"
    'keyC': 'ghi'  # will be ignored by BugsnagHandler
})

This is very useful if you are assigning context-specific attributes to your LogRecord objects, as described in the python logging cookbook.

Session tracking

BugSnag can track the number of “sessions” that happen within your application. This allows you to compare stability scores between releases and helps you to understand the quality of your releases.

You can start a session by calling bugsnag.start_session() at the start of an interval which should be considered a session.

Logging breadcrumbs

In order to understand what happened in your application before each error, it can be helpful to leave short log statements that we call breadcrumbs. A configurable number of breadcrumbs are attached to each error report to help diagnose what events led to the error.

Automatically captured breadcrumbs

By default, BugSnag captures the following events as breadcrumbs:

  • Error reports

This can be controlled using the enabled_breadcrumb_types configuration option.

Leaving manual breadcrumbs

You can use the leave_breadcrumb method to log potentially useful events in your own applications:

bugsnag.leave_breadcrumb("Something happened!")

Additional data can also be attached to breadcrumbs by providing the optional type and metadata parameters. For more information and examples for how custom breadcrumbs can be integrated, see customizing breadcrumbs.

Next steps

  • View bugsnag-python, the library powering this integration, on GitHub
  • Get support for your questions and feature requests