fix: set harakiri Django option to 30s (#3792)

* fix: set harakiri Django option to 30s

Closes #1573

From @guoard and @aminvakil who found the issue of "uWSGI reports full listen queue" was caused mostly by workers taking longer times to finish. For folks with bigger Sentry installation, they might want to increase the `proxy_read_timeout` and `harakiri` values to a longer (acceptable) time.

See the GitHub issue linked above for more details.

* feat: document 'harakiri' option instead of making it the default

* Update sentry.conf.example.py

Co-authored-by: Amin Vakil <[email protected]>

* Update sentry.conf.example.py

Co-authored-by: Amin Vakil <[email protected]>

---------

Co-authored-by: Amin Vakil <[email protected]>

Author: aldy505

sentry/sentry.conf.example.py

@@ -230,6 +230,12 @@ def get_internal_network():
     "workers": 3,
     "threads": 4,
     "memory-report": False,
+    # The `harakiri` option terminates requests that take longer than the
+    # defined amount of time (in seconds) which can help avoid stuck workers
+    # caused by GIL issues or deadlocks.
+    # Ensure nginx `proxy_read_timeout` configuration (default: 30)
+    # on your `nginx.conf` file to be at least 5 seconds longer than this.
+    # "harakiri": 25,
     # Some stuff so uwsgi will cycle workers sensibly
     "max-requests": 100000,
     "max-requests-delta": 500,