sensenet EVENTS

This backend feature is meant to serve as a foundation of different event handling features. The center of the Events feature is the repository, even outside events are directed through the repository service.

It consists of two parts: an Event Distributor and the Processors. Both run inside the sensenet application app domain.

Event Distributor

• content events (CRUD, versioning/approval events) are recorded
• when an event occurs, all processors are notified in parallel
• there are two types of processors:
  • synchronous: audit log, node observer
  • asynchronous: all new processors. These will run in the background, not blocking the content operation.

Later

• other events (permission change, login/logout)
• events are stored persistently
• changed data (metadata only) is stored, one per event
• called flags - if all set?, the event can be deleted (TBD)
• short life cycle, events are cleaned up regularly to save storage space
• the stored data is event-specific, but not related to the subscribers (observers), because the same event can be used by multiple different subscribers

Processor

• a processor examines the event and decides whether it is relevant or not
• processor types:
  • WEBHOOK
  • EMAIL (formerly Notification) (later)
  • SignalR (push notification for clients) (later)
• all processors get the same event data
• all processors get all events
• aggregating events is the responsibility of the processor (e.g. sending batch emails once a week)

Considerations

• the feature should not interfere with the events (does not block content save), runs in the background
• processing should be scalable
• new types of observers can be plugged in
• how and where do we store subscriptions? (TBD)
• filtering (for type, path, etc.) happens on the processor? (TBD)
• robustness and disaster recovery:
  • an event should be passed to an observer once and only once
  • if processing terminates before calling all observers, the processor should resume the calling after recovery

Some UML

Components and data

Collaboration