Sampling Events

You can enable event sampling to reduce the number of events sent to Honeybadger. This is especially useful if you are hitting your daily event quota limit:

python

from honeybadger import honeybadger

honeybadger.configure(events_sample_rate=50) # Sample 50% of events

The events_sample_rate option accepts a whole percentage value between 0 and 100, where 0 means no events will be sent and 100 means all events will be sent. The default is no sampling (100%).

Events are sampled by hashing the request_id if available in the event payload, otherwise a unique UUID is generated for that event. This deterministic approach ensures that related events from the same request are consistently sampled together.

Per-event sampling

Note

The _hb metadata is automatically removed from events before they are sent to Honeybadger.

The sample rate is applied to all events sent to Honeybadger Insights, including automatic instrumentation events. You can also set the sample rate per event by adding the _hb metadata with a sample_rate key:

python

  

honeybadger.event("user_created", {
    "user_id": user.id,
    "_hb": {"sample_rate": 100}
})

The event sample rate can also be set within the honeybadger.set_event_context function. This can be handy if you want to set an overall sample rate for a thread or ensure that specific instrumented events get sent:

python

# Set a higher sampling rate for this entire thread
honeybadger.set_event_context(_hb={"sample_rate": 100})

# Now all events from this thread, including automatic instrumentation,
# will use the 100% sample rate

Remember that event context is thread-local and applies to all events sent from the same thread after the set_event_context call, until it's changed or the thread terminates.

Sample rate precedence

Sample rates are determined in the following order (highest to lowest priority):

Event data - _hb metadata passed directly to the event
Event context - _hb metadata set via set_event_context()
Global configuration - events_sample_rate config option

When setting sample rates below the global setting, be aware that this affects how events with the same request_id are sampled. Since sampling is deterministic based on the request_id hash, all events sharing the same request_id will either all be sampled or all be skipped together. This ensures consistency across related events.