[Axon 3.1] What are the interfaces or abstract to implements a new Storage

Imported from Google Groups
1

Hi,

What are the interfaces/abstract classes to implement to create a new storage engine in Axon 3.1?

I saw they are possibly a lot of concepts to have the full support can you describe them ?

Thank you

Epo J.

2

Hello,
I am by no means an Axon expert, but until you get a better response, I thought that this image might help. My first instinct would be to look into extending BatchingEventStorageEngine. Of course, it really depends on your actual requirements, but that might make for a good starting point.

EventStorageEngine.PNG

3

Thank you Brian,

I thought there would have more classes/interfaces involved as it might include stuff for replaying events etc …

JDBC and JPA implementation I read somewhere needs 5 tables, can someone detail what is the purpose for each ?

Thank you

4

Hi Epo,

Below you find a short description for each table, hope this helps:

  • ASSOCIATION_VALUE_ENTRY: contains the associations for all Sagas, each row is one association for one saga instance (one saga can have multiple association values)
  • DOMAIN_EVENT_ENTRY: contains all serialized domain events, each row is one event
  • SAGA_ENTRY: contains all saga instances, each row is one saga
  • SNAPSHOT_EVENT_ENTRY: contains the snapshots for aggregates, each row is one snapshot
  • TOKEN_ENTRY: contains the state of the tracking event processors, a tracking event processor can have multiple segments, each row contains the administration for one segment of a processing group: tracking token, owner, etc. (see https://docs.axonframework.org/part3/event-processing.html#parallel-processing)

Cheers, Oscar

5

Hi Epo,

most of the logic you’d need for the Event Store is implemented in the Event Store. The EventStorageEngine is responsible for the actual storage of events. The BatchingEventStorageEngine is a useful subclass if the storage you use doesn’t know how to directly stream data. It assumes it needs to query a stream in batches, instead of opening a stream to the underlying storage directly.

As Oscar pointed out, you will only need to create the equivalents of the DomainEventEntry and SnapshotEventEntry tables in your storage.

Cheers,

Allard