Firing an event will place the event onto the bus and return immediately. No information is provided as to whether the event was processed, or indeed of it was received by any other service at all. No return value is provided when firing an event.

This is useful when:

  1. You wish to allow non-authoritative services to receive information without needing to concern yourself with their implementation
  2. You wish the authoritative service to perform a known task in the background

The quickstart provides an example of the latter case.

At-least-once semantics

Delivering a message exactly once is Very Difficult. Delivering a message at-most-once, or at-least-once is much more practical. Lightbus therefore provides at-least-once delivery for events.

As a result you can assume your event listeners will always receive an event, but sometimes a listener may be called multiple times for the same event.

You can handle this by ensuring your event listeners are idempotent. That is, implement your event listeners in such a way that it doesn't matter how many times they are executed.

See how to write idempotent event handlers.

Service names & listener names

An event will be delivered once to each consumer group. A consumer group is identified by a name in the form:

# Consumer group naming format

Your service name is specified in your service-level configuration. Your listener name is setup when you create your event listener (see below).

For example, this bus module sets up two listeners. Each listener is given a listener_name, thereby ensuring each listener receives a copy of every competitor_prices.changed event.

from my_handlers import send_price_alerts, update_db

bus = lightbus.create(

# Consumer group name: price-monitor-send-price-alerts

# Consumer group name: price-monitor-update-db


Failure to specify listener_name in the above example will result in each message going to either one listener or the other, but never to both. This is almost certainly not what you want.

Process names

Your service-level configuration also specifies a process name. This is less critical than the service name & listener name pairing described above, but still serves an important function.

The process name is used to track which process within a consumer group is dealing with a given event.

This is primarily useful when a service runs multiple Lightbus processes, normally as a result of scaling or reliability requirements. The purpose of a process name is twofold:

  1. Each process has a per-determined length of time to handle and acknowledge a given event. If this time is exceeded then failure will be assumed, and the event may be picked up by another process.
  2. When a Lightbus process starts up it will check for any outstanding events reserved for its process name. In which case it will process these messages first. This can happen if the process was killed prior to acknowledging messages it was processing.

Providing there is no timeout, an event will only be delivered to one process per listener within a service.

The default process name is a random 4 character string. If left unchanged, then clause 2 (above) will never be triggered (as a process name will not persist between process restarts). Any outstanding messages will always have to wait for the timeout period to expire, at which point the will be picked up by another process.


  • More complex
  • More robust

Best practices

Event naming

  • Past tense
  • Domain-based vs technical-based

Parameter values

  • Consistent meaning over time (not 'tomorrow', or '6 days ago')
  • Chop up your relations (aggregates in DDD)