SENTRY 2.0 (this) IS IN DEVELOPMENT AND SHOULD NOT BE USED IN PRODUCTION
Sentry provides you with a generic interface to view and interact with your error logs. By default, it will record various events to a datastore. With this it allows you to interact and view near real-time information to discover issues and more easily trace them in your application.
(The next chunk is a lie, but planned)
Built-in support includes:
- Drop-in Django support
- WSGI middleware for error logging
- Query logging for psycopg2 and MySQLdb
- logging and logbook modules
Issue tracker: http://github.com/dcramer/django-sentry/issues
We'll move all of this into the Sphinx docs once APIs are finalized.
Sentry uses Nose, which will automatically be installed (along with unittest2) if you use the test command.
mkvirtualenv sentry python setup.py test
The server component of Sentry, called sentry.web, can be run with the following command:
mkvirtualenv sentry python setup.py develop sentry start --no-daemon --debug
The client is the core of Sentry, which is composed of the sentry namespace, and its many submodules. There are a two key things that make up the client:
These components are also used within sentry.web as well.
To capture an error, Sentry provides the capture function:
from sentry import capture capture('sentry.events.Exception', exc_info=exc_info)
For built-in events, you can drop the module namespace:
from sentry import capture capture('Exception', exc_info=exc_info)
One of the big changes from Sentry 1.x to 2.x, is the ability to record durations (in ms) with events:
from sentry import capture capture('Query', query='SELECT * FROM table', engine='psycopg2', time_spent=36)
The other major change, is that labels like "logger", and "server" are now part of the tagging architecture:
from sentry import capture capture('Message', message='hello world', tags=[('logger', 'root'), ('level', 'error'), ('url', 'http://example.com')])
The filter API is designed to allow dynamic filters based on tags. They are composed of two pieces: a processor and a renderer. The renderer simply tells Sentry how it needs to be displayed in the filter list (e.g. a select widget with FOO choices, a search input, etc).
The majority of this code lies within sentry.web.filters, and are specified as part of the SLICES runtime configuration.
This is a rough list of features/APIs which need to be completed (some of this is better described in the issue tracker):
- Runtime validation of settings (ensure events and filters are valid importables, etc.)
- Exception event handler needs to have an intelligent hash
- Interfaces may be able to be mixed somehow with Events (as they both provide serialization). Possibly have events implement an interface?
- Processor API (for post-processing messages before they get sent to the server)
- Should be an ordered tuple specified in config
- Should be able to manipulate all of data (so you can clean messages, or add additional metadata)
- Filter API
- Needs to handle sorting+filtering with denormalized indexes
- Plugin API
- Dashboard view needs finalized
- Need to properly index/query on sort+tag combinations
- Implement pagination
- Should consider supporting better interval dashboards. e.g. last 24 hours, vs last 15 minutes
- Django Integration (some draft code is present)
- SQLAlchemy Backend
- Test coverage
- Models and Backends
- Decide on final version of client authentication API
- Probably don't need it to be so secure (nonce is extra load)
- Support should be considered for having multiple "users". A good example use case is if a consulting firm uses a single Sentry server and has many clients, but then decides one client's access needs revoked from recording to the logger.
- Make reporting extendable
- The email reporting which was available in Sentry 1.x should simply be a builtin reporting option.
- Add an IRC extension?
- Add a network Growl extension?
- Add an IM extension?