Advanced Usage

This covers some advanced usage scenarios for raven Python.

If you want to use the latest git version you can get it from the GitHub repository:

Copied
git clone https://github.com/getsentry/raven-python
pip install raven-python

Certain additional features can be installed by defining the feature when pip installing it. For instance to install all dependencies needed to use the Flask integration, you can depend on raven[flask]:

Copied
pip install raven[flask]

For more information refer to the individual integration documentation.

Settings are specified as part of the initialization of the client. The client is a class that can be instantiated with a specific configuration and all reporting can then happen from the instance of that object. Typically an instance is created somewhere globally and then imported as necessary.

Copied
from raven import Client

# Read configuration from the ``SENTRY_DSN`` environment variable
client = Client()

# Manually specify a DSN
client = Client('https://examplePublicKey@o0.ingest.sentry.io/0')

A reasonably configured client should generally include a few additional settings:

Copied
import os
import raven

client = raven.Client(
    dsn='https://examplePublicKey@o0.ingest.sentry.io/0',

    # inform the client which parts of code are yours
    # include_paths=['my.app']
    include_paths=[__name__.split('.', 1)[0]],

    # pass along the version of your application
    # release='my-project-name@2.3.12'
    # release=raven.fetch_package_version('my-app')
    release=raven.fetch_git_sha(os.path.dirname(__file__)),
)

(New in version 5.2.0: The fetch_package_version and fetch_git_sha helpers.)

The following are valid arguments which may be passed to the Raven client:

dsn

A Sentry compatible DSN as mentioned before:

Copied
dsn = 'https://examplePublicKey@o0.ingest.sentry.io/0'

transport

The HTTP transport class to use. By default this is an asynchronous worker thread that runs in-process.

For more information see Transports.

site

An optional, arbitrary string to identify this client installation:

Copied
site = 'my site name'

name

This will override the server_name value for this installation. Defaults to socket.gethostname():

Copied
name = 'sentry_rocks_' + socket.gethostname()

release

The version of your application. This will map up into a Release in Sentry:

Copied
release = '1.0.3'

environment

The environment your application is running in:

Copied
environment = 'staging'

tags

Default tags to send with events:

Copied
tags = {'site': 'foo.com'}

exclude_paths

Extending this allow you to ignore module prefixes when we attempt to discover which function an error comes from (typically a view):

Copied
exclude_paths = [
    'django',
    'sentry',
    'raven',
    'lxml.objectify',
]

include_paths

For example, in Django this defaults to your list of INSTALLED_APPS, and is used for drilling down where an exception is located:

Copied
include_paths = [
    'django',
    'sentry',
    'raven',
    'lxml.objectify',
]

ignore_exceptions

A list of exceptions to ignore:

Copied
ignore_exceptions = [
    'Http404',
    'django.exceptions.http.Http404',
    'django.exceptions.*',
    ValueError,
]

Each item can be either a string or a class. String declaration is strict (ie. does not work for child exceptions) whereas class declaration handle inheritance (ie. child exceptions are also ignored).

sample_rate

The sampling factor to apply to events. A value of 0.00 will deny sending any events, and a value of 1.00 will send 100% of events.

Copied
# send 50% of events
sample_rate = 0.5

list_max_length

The maximum number of items a list-like container should store.

If an iterable is longer than the specified length, the left-most elements up to length will be kept. This affects sets as well, which are unordered.

Copied
list_max_length = 50

string_max_length

The maximum characters of a string that should be stored.

If a string is longer than the given length, it will be truncated down to the specified size:

Copied
string_max_length = 200

auto_log_stacks

Should Raven automatically log frame stacks (including locals) for all calls as it would for exceptions:

Copied
auto_log_stacks = True

processors

A list of processors to apply to events before sending them to the Sentry server. Useful for sending additional global state data or sanitizing data that you want to keep off of the server:

Copied
processors = (
    'raven.processors.SanitizePasswordsProcessor',
)

sanitize_keys

A required set of keys to sanitize when using raven.processors.SanitizeKeysProcessor in processors:

Copied
sanitize_keys = ["sensitive_key_1", "sensitive_key_2"]

Several processors are included with Raven to assist in data sanitization. These are configured with the processors value.

raven.processors.SanitizePasswordsProcessor

Removes all keys which resemble password, secret, or api_key within stack trace contexts, HTTP bits (such as cookies, POST data, the querystring, and environment), and extra data.

raven.processors.RemoveStackLocalsProcessor

Removes all stack trace context variables. This will cripple the functionality of Sentry, as you’ll only get raw tracebacks, but it will ensure no local scoped information is available to the server.

raven.processors.RemovePostDataProcessor

Removes the body of all HTTP data.

raven.processors.SanitizeKeysProcessor

Removes all keys provided in the configuable set sanitize_keys, which must be provided as a client argument.

In some cases you may see issues where Sentry groups multiple events together when they should be separate entities. In other cases, Sentry simply doesn’t group events together because they’re so sporadic that they never look the same.

Both of these problems can be addressed by specifying the fingerprint attribute.

For example, if you have HTTP 404 (page not found) errors, and you’d prefer they deduplicate by taking into account the URL:

Copied
client.captureException(fingerprint=['{{ default }}', 'http://my-url/'])

For more information, see Customize Grouping with Fingerprints.

There are two ways to sample messages:

  • Add sample_rate to the Client object - This sends a percentage of messages the reaching the Client to Sentry
Copied
client = Client('https://examplePublicKey@o0.ingest.sentry.io/0', sample_rate=0.5) # send 50% of events
  • Sample individual messages
Copied
client = Client('https://examplePublicKey@o0.ingest.sentry.io/0') # No sample_rate provided

try:
    1 / 0
except ZeroDivisionError:
    client.captureException(sample_rate=0.5) # Send 50% of this event

Alternatively, if you have SentryHandler configured in your logging stack, you can send sample_rate in the extra kwarg in each log like this

Copied
some_logger.warning('foo', extra={'sample_rate': 0.5}) # Send 50% of this event

If you’re using uWSGI you will need to add enable-threads to the default invocation, or you will need to switch off of the threaded default transport.

Help improve this content
Our documentation is open source and available on GitHub. Your contributions are welcome, whether fixing a typo (drat!) or suggesting an update ("yeah, this would be better").