rollbar

Easy and powerful exception tracking with Rollbar. Send messages and exceptions with arbitrary context, get back aggregates, and debug production issues quickly.

Project Description

Python notifier for reporting exceptions, errors, and log messages to Rollbar.

Quick start

Install using pip:

pip install rollbar

import rollbar
rollbar.init('POST_SERVER_ITEM_ACCESS_TOKEN', 'production')  # access_token, environment

try:
    main_app_loop()
except IOError:
    rollbar.report_message('Got an IOError in the main loop', 'warning')
except:
    # catch-all
    rollbar.report_exc_info()
    # equivalent to rollbar.report_exc_info(sys.exc_info())

Requirements

Python 2.7, 3.3, 3.4, or 3.5
requests 0.12+
A Rollbar account

Configuration

Django

In your settings.py, add 'rollbar.contrib.django.middleware.RollbarNotifierMiddleware' as the last item in

MIDDLEWARE_CLASSES in Django 1.9 and earlier:

python MIDDLEWARE_CLASSES = [ # ... other middleware classes ... 'rollbar.contrib.django.middleware.RollbarNotifierMiddleware', ]

MIDDLEWARE in Django 1.10 and up:

python MIDDLEWARE = [ # ... other middleware classes ... 'rollbar.contrib.django.middleware.RollbarNotifierMiddleware', ]

Add these configuration variables in settings.py:

ROLLBAR = {
    'access_token': 'POST_SERVER_ITEM_ACCESS_TOKEN',
    'environment': 'development' if DEBUG else 'production',
    'branch': 'master',
    'root': '/absolute/path/to/code/root',
}

Be sure to replace POST_SERVER_ITEM_ACCESS_TOKEN with your project’s post_server_item access token, which you can find in the Rollbar.com interface.

Check out the Django example.

If you’d like to be able to use a Django LOGGING handler that could catch errors that happen outside of the middleware and ship them to Rollbar, such as in celery job queue tasks that run in the background separate from web requests, do the following:

Add this to the handlers key:

'rollbar': {
    'filters': ['require_debug_false'],
    'access_token': 'POST_SERVER_ITEM_ACCESS_TOKEN',
    'environment': 'production',
    'class': 'rollbar.logger.RollbarHandler'
},

Then add the handler to the loggers key values where you want it to fire off.

'myappwithtasks': {
    'handlers': ['console', 'logfile', 'rollbar'],
    'level': 'DEBUG',
    'propagate': True,
},

Pyramid

In your ini file (e.g. production.ini), add rollbar.contrib.pyramid to the end of your pyramid.includes:

[app:main]
pyramid.includes =
    pyramid_debugtoolbar
    rollbar.contrib.pyramid

And add these rollbar configuration variables:

[app:main]
rollbar.access_token = POST_SERVER_ITEM_ACCESS_TOKEN
rollbar.environment = production
rollbar.branch = master
rollbar.root = %(here)s

Be sure to replace POST_SERVER_ITEM_ACCESS_TOKEN with your project’s post_server_item access token, which you can find in the Rollbar.com interface.

The above will configure Rollbar to catch and report all exceptions that occur inside your Pyramid app. However, in order to catch exceptions in middlewares or in Pyramid itself, you will also need to wrap your app inside a pipeline with Rollbar as a filter.

To do this, first change your ini file to use a pipeline. Change this:

[app:main]
#...

To:

[pipeline:main]
pipeline =
    rollbar
    YOUR_APP_NAME

[app:YOUR_APP_NAME]
pyramid.includes =
    pyramid_debugtoolbar
    rollbar.contrib.pyramid

rollbar.access_token = POST_SERVER_ITEM_ACCESS_TOKEN
rollbar.environment = production
rollbar.branch = master
rollbar.root = %(here)s

[filter:rollbar]
use = egg:rollbar#pyramid
access_token = POST_SERVER_ITEM_ACCESS_TOKEN
environment = production
branch = master
root = %(here)s

Note that the access_token, environment, and other Rollbar config params do need to be present in both the app section and the filter section.

Additionally, note that because Pyramid uses INI files for configuration, any changes to nested settings, like the locals dictionary, will need to be handled in code.

Flask

Check out rollbar-flask-example.

Be sure to add the required blinker dependency! See requirements.txt in the example repo for how.

Bottle

Import the plugin and install! Can be installed globally or on a per route basis.

import bottle
from rollbar.contrib.bottle import RollbarBottleReporter

rbr = RollbarBottleReporter(access_token='POST_SERVER_ITEM_ACCESS_TOKEN', environment='production') #setup rollbar

bottle.install(rbr) #install globally

@bottle.get('/')
def raise_error():
  '''
  When navigating to /, we'll get a regular 500 page from bottle,
  as well as have the error below listed on Rollbar.
  '''
  raise Exception('Hello, Rollbar!')

if __name__ == '__main__':
    bottle.run(host='localhost', port=8080)

Be sure to replace POST_SERVER_ITEM_ACCESS_TOKEN with your project’s post_server_item access token, which you can find in the Rollbar.com interface.

Twisted

Check out the Twisted example.

AWS Lambda

The biggest issue with the Lambda execution environment is that as soon as you return from your handler function, any work executing in other threads will stop executing as the process is frozen. This is true also of any child processes that one may spawn. Furthermore, the Lambda environment implements multithreading via a hypervisor on a single CPU core. Therefore, using separate threads to do additional work will not necessarily lead to better performance.

In order to ensure that the Rollbar library works correctly, meaning that items are transmitted to the Rollbar API, one must not return from the main handler function before all of this work completes. In order to ensure this, one can either use the blocking handler by specifying this value in the configuration,

rollbar.init(token, environment='production', handler='blocking')

or use the Rollbar function wait to delay the return from your function until all Rollbar threads have finished. Note that we use threads for the handler if otherwise unspecified, therefore you must use wait if you do not set the handler.

wait is a function which takes an optional function as an argument. It waits for all currently running Rollbar created threads to stop processing, meaning it waits for any items to be sent over the network, then it returns the result of calling the function passed as an argument or None if function was given. Hence, one can use it via

def lambda_handler(event, context):
    try:
        result = ...
        return rollbar.wait(lambda: result)
    except:
        rollbar.report_exc_info()
        rollbar.wait()
        raise

We provide a decorator for your handler functions which takes care of calling wait properly as well as catching any exceptions, namely rollbar.lambda_function:

import os
import rollbar

token = os.getenv('ROLLBAR_KEY', 'missing_api_key')
rollbar.init(token, 'production')

@rollbar.lambda_function
def lambda_handler(event, context):
    return some_other_function('Hello from Lambda')

Other

For generic Python or a non-Django/non-Pyramid framework just initialize the Rollbar library with your access token and environment.

rollbar.init('POST_SERVER_ITEM_ACCESS_TOKEN', environment='production', **other_config_params)

Other options can be passed as keyword arguments. See the reference below for all options.

Command-line usage

pyrollbar comes with a command-line tool that can be used with other UNIX utilities to create an ad-hoc monitoring solution.

e.g. Report all 5xx haproxy requests as warning

tail -f /var/log/haproxy.log | awk '{print $11,$0}' | grep '^5' | awk '{$1="";print "warning",$0}' | rollbar -t POST_SERVER_ITEM_ACCESS_TOKEN -e production -v

e.g. Test an access token

rollbar -t POST_SERVER_ITEM_ACCESS_TOKEN -e test debug testing access token

Reference

$ rollbar --help
Usage: rollbar [options]

Options:
  --version             show program's version number and exit
  -h, --help            show this help message and exit
  -t ACCESS_TOKEN, --access_token=ACCESS_TOKEN
                        You project's access token from rollbar.com.
  -e ENVIRONMENT, --environment=ENVIRONMENT
                        The environment to report errors and messages to.
  -u ENDPOINT_URL, --url=ENDPOINT_URL
                        The Rollbar API endpoint url to send data to.
  -m HANDLER, --handler=HANDLER
                        The method in which to report errors.
  -v, --verbose         Print verbose output.

Usage

The Django, Pyramid, Flask, and Bottle integrations will automatically report uncaught exceptions to Rollbar.

Exceptions

To report a caught exception to Rollbar, use rollbar.report_exc_info():

try:
    do_something()
except:
    rollbar.report_exc_info(sys.exc_info())
    # or if you have a webob-like request object, pass that as well:
    # rollbar.report_exc_info(sys.exc_info(), request)

Logging

You can also send any other log messages you want, using rollbar.report_message():

try:
    do_something()
except IOError:
    rollbar.report_message('Got an IOError while trying to do_something()', 'warning')
    # report_message() also accepts a request object:
    #rollbar.report_message('message here', 'warning', request)

Examples

Here’s a full example, integrating into a simple Gevent app.

"""
Sample Gevent application with Rollbar integration.
"""
import sys
import logging

from gevent.pywsgi import WSGIServer
import rollbar
import webob

# configure logging so that rollbar's log messages will appear
logging.basicConfig()

def application(environ, start_response):
    request = webob.Request(environ)
    status = '200 OK'
    headers = [('Content-Type', 'text/html')]
    start_response(status, headers)

    yield '<p>Hello world</p>'

    # extra fields we'd like to send along to rollbar (optional)
    extra_data = {'datacenter': 'us1', 'app' : {'version': '1.1'}}

    try:
        # will raise a NameError about 'bar' not being defined
        foo = bar
    except:
        # report full exception info
        rollbar.report_exc_info(sys.exc_info(), request, extra_data=extra_data)

        # and/or, just send a string message with a level
        rollbar.report_message("Here's a message", 'info', request, extra_data=extra_data)

        yield '<p>Caught an exception</p>'

# initialize rollbar with an access token and environment name
rollbar.init('POST_SERVER_ITEM_ACCESS_TOKEN', 'development')

# now start the wsgi server
WSGIServer(('', 8000), application).serve_forever()

Configuration reference

access_token

Access token from your Rollbar project

agent.log_file

If handler is agent, the path to the log file. Filename must end in .rollbar

branch

Name of the checked-out branch.

Default: master

code_version

A string describing the current code revision/version (i.e. a git sha). Max 40 characters.

Default: None

enabled

Controls whether or not Rollbar will report any data

Default: True

endpoint

URL items are posted to.

Default: https://api.rollbar.com/api/1/item/

environment

Environment name. Any string up to 255 chars is OK. For best results, use “production” for your production environment.

exception_level_filters

List of tuples in the form (class, level) where class is an Exception class you want to always filter to the respective level. Any subclasses of the given class will also be matched.

Valid levels: 'critical', 'error', 'warning', 'info', 'debug' and 'ignored'.

Use 'ignored' if you want an Exception (sub)class to never be reported to Rollbar.

Any exceptions not found in this configuration setting will default to 'error'.

Django settings.py example (and Django default):
from django.http import Http404

ROLLBAR = {
    ...
    'exception_level_filters': [
        (Http404, 'warning')
    ]
}
In a Pyramid ini file, define each tuple as an individual whitespace delimited line, for example:
rollbar.exception_level_filters =
    pyramid.exceptions.ConfigurationError critical
    #...

handler

The method for reporting rollbar items to api.rollbar.com

One of:

blocking – runs in main thread
thread – spawns a new thread
agent – writes messages to a log file for consumption by rollbar-agent
tornado – uses the Tornado async library to send the payload
gae – uses the Google AppEngineFetch library to send the payload
twisted – uses the Twisted event-driven networking library to send the payload

Default: thread

locals

Configuration for collecting local variables. A dictionary:

enabled: If True, variable values will be collected for stack traces. Default True.
safe_repr: If True, non-built-in objects will be serialized into just their class name. If False repr(obj) will be used for serialization. Default True.

sizes

Dictionary of configuration describing the max size to repr() for each type.

maxdict: Default 10
maxarray: Default 10
maxlist: Default 10
maxtuple: Default 10
maxset: Default 10
maxfrozenset: Default 10
maxdeque: Default 10
maxstring: Default 100
maxlong: Default 40
maxother: Default 100

whitelisted_types

A list of type objects, (e.g. type(my_class_instance) or MyClass) that will be serialized using repr(). Default []

scrub_varargs

If True, variable argument values will be scrubbed. Default True.

root

Absolute path to the root of your application, not including the final /.

scrub_fields

List of sensitive field names to scrub out of request params and locals. Values will be replaced with asterisks. If overriding, make sure to list all fields you want to scrub, not just fields you want to add to the default. Param names are converted to lowercase before comparing against the scrub list.

Default: ['pw', 'passwd', 'password', 'secret', 'confirm_password', 'confirmPassword', 'password_confirmation', 'passwordConfirmation', 'access_token', 'accessToken', 'auth', 'authentication']

timeout

Timeout for any HTTP requests made to the Rollbar API (in seconds).

Default: 3

allow_logging_basic_config

When True, logging.basicConfig() will be called to set up the logging system. Set to False to skip this call. If using Flask, you’ll want to set to False. If using Pyramid or Django, True should be fine.

Default: True

url_fields

List of fields treated as URLs and scrubbed. Default ['url', 'link', 'href']

verify_https

If True, network requests will fail unless encountering a valid certificate. Default True.

shortener_keys

A list of key prefixes (as tuple) to apply our shortener transform to.

Added to built-in list:

[
    ('body', 'request', 'POST'),
    ('body', 'request', 'json')
]

If locals.enabled is True, extra keys are also automatically added:

[
    ('body', 'trace', 'frames', '*', 'code'),
    ('body', 'trace', 'frames', '*', 'args', '*'),
    ('body', 'trace', 'frames', '*', 'kwargs', '*'),
    ('body', 'trace', 'frames', '*', 'locals', '*')
]

Default: []

suppress_reinit_warning