Replaying events using snapshot strategy

Gustavo_Monti · February 22, 2019, 7:22pm

Hi,

I’m able to create snapshots every X events. After deleting token and entire database the axon starts to replaying events (I’m doing this way because i’m testing this). What I’m seeing is that all events (from start to past) is being replayed generating a large amount of events. How can I replay events considering the snaphost strategy ?

Ex:

Snapshoting

Event 1
Event 2
Event 3
----------------- Snapshot (every 3 events)
Event 4
Event 5

Replaying Events

Event 4
Event 5

Thanks,

Matt_Newman · February 26, 2019, 9:33pm

As you said you delete the tokens, I guess this implies that you’re using a TrackingEventProcessor to feed events to one or more handlers.

The event processors are (AFAIK) not aware of snapshots, as a snapshot only makes sense when looking at an individual aggregate. Event processors deal with all messages across all aggregates. This is powerful as some read models which are being built from the TrackingEventProcessor may need more information than what’s available in a snapshot.

If you were to load the aggregate which has been snapshotted using a Repository, you should see that only the snapshot and subsequent events are read. When dispatching commands, Axon loads aggregates in this manner. You can manually do it by injecting the appropriate repo and calling Repository#load(String).

If you want to dig into the source code:

The event processors use a StreamableMessageSource, which per the documentation returns ‘a stream containing all messages’
The EventSourcingRepository uses one of the EventStore#readEvents(…) overloads that accepts an aggregateIdentifier, which return only events for a particular aggregate

HTH!

Cheers,
Matt