django-watchman exposes a status endpoint for your backing services like databases, caches, etc.
The full documentation is at http://django-watchman.rtfd.org.
We're in love with django-watchman. External monitoring is a vital part of our service offering. Using django-watchman we can introspect the infrastructure of an application via a secure URL. It's very well written and easy to extend. We've recommended it to many of our clients already.
— Hany Fahim, CEO, VM Farms.
Install
django-watchman
:pip install django-watchman
Add
watchman
to yourINSTALLED_APPS
setting like this:INSTALLED_APPS = ( ... 'watchman', )
Include the watchman URLconf in your project
urls.py
like this:url(r'^watchman/', include('watchman.urls')),
Start the development server and visit
http://127.0.0.1:8000/watchman/
to get a JSON response of your backing service statuses:{ "databases": [ { "default": { "ok": true } } ], "caches": [ { "default": { "ok": true } } ], "storage": {"ok": true} }
Visit http://127.0.0.1:8000/watchman/dashboard/
to get a human-friendly HTML
representation of all of your watchman checks.
If you want to protect the status endpoint, you can use the WATCHMAN_TOKENS
setting.
This is a comma-separated list of tokens.
When this setting is added, you must pass one of the tokens in as the watchman-token
GET parameter:
GET http://127.0.0.1:8000/watchman/?watchman-token=:token
Or by setting the Authorization: WATCHMAN-TOKEN
header on the request:
curl -X GET -H "Authorization: WATCHMAN-TOKEN Token=\":token\"" http://127.0.0.1:8000/watchman/
If you want to change the token name, you can set the WATCHMAN_TOKEN_NAME
.
The value of this setting will be the GET parameter that you must pass in:
WATCHMAN_TOKEN_NAME = 'custom-token-name' GET http://127.0.0.1:8000/watchman/?custom-token-name=:token
DEPRECATION WARNING: WATCHMAN_TOKEN
was replaced by the WATCHMAN_TOKENS
setting to support multiple authentication tokens in django-watchman 0.11
.
It will continue to work until it's removed in django-watchman 1.0
.
If you want to protect the status endpoint with a customized
authentication/authorization decorator, you can add WATCHMAN_AUTH_DECORATOR
to your settings. This needs to be a dotted-path to a decorator, and defaults
to watchman.decorators.token_required
:
WATCHMAN_AUTH_DECORATOR = 'django.contrib.admin.views.decorators.staff_member_required'
Note that the token_required
decorator does not protect a view unless
WATCHMAN_TOKENS
is set in settings.
django-watchman allows you to customize the checks which are run by modifying
the WATCHMAN_CHECKS
setting. In settings.py
:
WATCHMAN_CHECKS = ( 'module.path.to.callable', 'another.module.path.to.callable', )
You can also import the watchman.constants to include the DEFAULT_CHECKS and PAID_CHECKS in your settings.py
:
from watchman import constants as watchman_constants WATCHMAN_CHECKS = watchman_constants.DEFAULT_CHECKS + ('module.path.to.callable', )
Checks take no arguments, and must return a dict
whose keys are applied to the JSON response. Use the watchman.decorators.check
decorator to capture exceptions:
from watchman.decorators import check @check def my_check(): return {'x': 1}
In the absence of any checks, a 404 is thrown, which is then handled by the
json_view
decorator.
A subset of checks may be run, by passing ?check=module.path.to.callable&check=...
in the request URL. Only the callables given in the querystring which are also
in WATCHMAN_CHECKS
should be run, eg:
curl -XGET http://127.0.0.1:8080/watchman/?check=watchman.checks.caches
You can skip any number of checks, by passing ?skip=module.path.to.callable&skip=...
in the request URL. Only the checks in WATCHMAN_CHECKS
which are not in the
querystring should be run, eg:
curl -XGET http://127.0.0.1:8080/watchman/?skip=watchman.checks.email
If your application has a large number of databases or caches configured, watchman may open too many connections as it checks each database or cache.
You can set the WATCHMAN_DATABASES
or WATCHMAN_CACHES
settings in order
to override the default set of databases and caches to be monitored.
If you want to simply check that your application is running and able to handle requests, you can call ping:
GET http://127.0.0.1:8000/watchman/ping/
It will return the text pong
with a 200 status code. Calling this doesn't
run any of the checks.
If you would like a "bare" status view (one that doesn't report any details,
just HTTP 200
if checks pass, and HTTP 500
if any checks fail), you
can use the bare_status
view by putting the following into urls.py
:
import watchman.views # ... url(r'^status/?$', watchman.views.bare_status),
You can also run your checks without starting the webserver and making requests. This can be useful for testing your configuration before enabling a server, checking configuration on worker servers, etc. Run the management command like so:
python manage.py watchman
By default, successful checks will not print any output. If all checks pass
successfully, the exit code will be 0
. If a check fails, the exit code will
be 1
, and the error message including stack trace will be printed to stderr
.
If you'd like to see output for successful checks as well, set verbosity to
2
or higher:
python manage.py watchman -v 2 {"storage": {"ok": true}} {"caches": [{"default": {"ok": true}}]} {"databases": [{"default": {"ok": true}}]}
If you'd like to run a subset of checks, use -c
and a comma-separated list
of python module paths:
python manage.py watchman -c watchman.checks.caches,watchman.checks.databases -v 2 {"caches": [{"default": {"ok": true}}]} {"databases": [{"default": {"ok": true}}]}
If you'd like to skip certain checks, use -s
and a comma-separated list of
python module paths:
python manage.py watchman -s watchman.checks.caches,watchman.checks.databases -v 2 {"storage": {"ok": true}}
Use -h
to see a full list of options:
python manage.py watchman -h
Watchman can return the version of watchman which is running to help you keep
track of whether or not your sites are using an up-to-date version. This is
disabled by default to prevent any unintended information leakage for websites
without authentication. To enable, update the EXPOSE_WATCHMAN_VERSION
setting:
EXPOSE_WATCHMAN_VERSION = True
By default, watchman will return a 500
HTTP response code, even if there's a
failing check. You can specify a different response code for failing checks
using the WATCHMAN_ERROR_CODE
setting:
WATCHMAN_ERROR_CODE = 200
watchman includes log messages using a logger called watchman
.
You can configure this by configuring the LOGGING
section of your Django
settings file.
Here is a simple example that would log to the console:
LOGGING = { 'version': 1, 'disable_existing_loggers': False, 'handlers': { 'console': { 'class': 'logging.StreamHandler', }, }, 'loggers': { 'watchman': { 'handlers': ['console'], 'level': 'DEBUG', }, }, }
More information is available in the Django documentation.
If you're using APM and watchman is being often hit for health checks (such as an ELB on AWS), you will find some stats based on averages will be affected (average transaction time, apdex, etc):
You can disable APM instrumentation for watchman by using the WATCHMAN_DISABLE_APM
setting:
WATCHMAN_DISABLE_APM = True
This currently supports the following agents:
- Datadog
- New Relic
Please open an issue if there's another APM you use which is being affected.
For each cache in django.conf.settings.CACHES
:
- Set a test cache item
- Get test item
- Delete test item
For each database in django.conf.settings.DATABASES
:
- Verify connection by calling
connections[database].introspection.table_names()
Send a test email to [email protected]
using django.core.mail.send_mail
.
If you're using a 3rd party mail provider, this check could end up costing you money, depending how aggressive you are with your monitoring. For this reason, this check is not enabled by default.
For reference, if you were using Mandrill, and hitting your watchman endpoint once per minute, this would cost you ~$5.60/month.
Custom Settings
WATCHMAN_EMAIL_SENDER
(default:[email protected]
): Specify an email to be the sender of the test emailWATCHMAN_EMAIL_RECIPIENTS
(default:[[email protected]]
): Specify a list of email addresses to send the test emailWATCHMAN_EMAIL_HEADERS
(default:{}
): Specify a dict of custom headers to be added to the test email
Using django.core.files.storage.default_storage
:
- Write a test file
- Check the test file's size
- Read the test file's contents
- Delete the test file
By default the test file gets written on the root of the django MEDIA_ROOT
. If for whatever reasons this path is not writable by the user that runs the application you can override it by setting WATCHMAN_STORAGE_PATH
to a specific path.
Remember that this must be within the MEDIA_ROOT
, which by default is your project root. In settings.py
:
WATCHMAN_STORAGE_PATH = "/path_to_your_app/foo/bar/"
If the MEDIA_ROOT
is already defined:
from os.path import join as joinpath WATCHMAN_STORAGE_PATH = joinpath(MEDIA_ROOT, "foo/bar")
By default, django-watchman will run checks against your databases
(watchman.checks.databases
), caches (watchman.checks.caches
), and
storage (watchman.checks.storage
).
Currently there is only one "paid" check - watchman.checks.email
. You can
enable it by setting the WATCHMAN_ENABLE_PAID_CHECKS
to True
, or by
overriding the WATCHMAN_CHECKS
setting.
A sample project is available along with a Dockerfile to make it easy to try out django-watchman.
- Docker <https://www.docker.com/get-docker>
- Build and run the Docker image with the current local code:
make run
- Visit watchman json endpoint in your browser: http://127.0.0.1:8000/watchman/
- Visit watchman dashboard in your browser: http://127.0.0.1:8000/watchman/dashboard/
- Visit watchman ping in your browser: http://127.0.0.1:8000/watchman/ping/
- Visit watchman bare status in your browser: http://127.0.0.1:8000/watchman/bare/