Event Sourcing with ESE

ESE stands for Event Sourced Entities, and is part of project Goodees, which is a collection of Java EE utilities.

We started ESE since we missed a programming model that integrates well with Java EE and offers entities with event sourced state and also serve for executing business methods. Think of it as "What if Entity EJBs were event sourced rather than ORM?"

Transcript

1. Event Sourcing with ESE Patrik Duditš
2. Event Sourcing with ESE Patrik Duditš Java EE developer Contributor to Payara @pdudits
3. Part of Goodees project • Utility classes for Java EE • https://github.com/goodees/goodees • Event Sourced Entities
4. Event Sourcing
5. Buzzword bingo sheet Aggregate root Bounded Context Sharding CQRS Asynchronous Actor model Microservices Resilient Kafka Cloud native Scalable Schemaless Reactive Message passing Event store Event storming Replication Domain Driven Process manager Distributed High availability Lock-free CQS Eventual Consistency Stream processing
6. CRUD systems • Structure first • Constraints • Database migrations
7. “Why your system acted like that two weeks ago?” “Let’s see the logs”
8. The idea of event sourcing You only store facts about past events. State can be reconstructed by replaying these events in order. All state in your application is transient. Wish List Bounded Context: Wish registered Parent registered Wish content changed Parent verified Gift bought Child registered Wish deleted
9. Wish List Bounded Context: Wish registered Parent registered Wish content changed Parent verified Gift bought Child registered Wish deleted Q?
10. Java Frameworks for event sourcing • Akka, Lagom • Axon Framework • Es4j • ESE – Event sourced Entities
11. Motivation We are aware of “Not invented here syndrome”, but we use our own, better fitting term for it.
12. Motivation Small Library – simpler API, quick to learn Easy integration with Java EE
13. Protocol Gateway Metrics & Alerting Charging & Roaming Mass Configuration Billing Charge Session Management Mobile app B2B APIs Roaming platforms SAP Electricity grid ~ 40 commands ~ 40 interfaces
14. Problems we faced • State of station scattered across subsystems • Implementation details of station protocols as well • Concurrent communication exchanges towards/from stations • Constructs similar to synchronize(station) {} • State synchronization in high availability setups
15. Features of ESE • Java 8 SE – few dependencies (slf4j, jackson) • Straight API – No annotations or reflection • Lock-free concurrency • Support for JSON serialization of events • Support for storing events via JDBC • Multiple execution modes (sync, async)
16. ESE Building blocks - Concepts Runtime • Access point for passing a request to specific entity • Integration with surrounding environment • Singleton Entity • Stateful object with an ID • Executes requests • Performs business logic • Not JPA entity • State built exclusively by applying events
17. ESE Building blocks - APIs Dispatcher • Guarantees one request per entity handled • Executes requests on a thread pool • Takes care of timeouts and retries Event sourcing APIs • API for entity lifecycle • Events • Persisting and application of events
18. Dispatching client request
19. Execution within entity "Jane wants a 100€ lego" @state: Is Jane registered in this wishlist? @behavior service: Was she obeyant enough?
20. Execution within entity event = new WishRegisteredEvent(128, “Jane”, “a 100€ lego”) eventStore.persist(event)
21. Execution within entity
22. Execution within entity
23. Execution within entity updateState(event) wishes.add(new Wish(event.getId(), event.getChild(), event.getContent()));
24. Execution within entity
25. Execution within entity Accepted, wishId = 128
26. Execution within entity “Buy wish id 128" @state: Is there such wish?
27. Execution within entity event = new PurchaseRequested(128) eventStore.persist(event)
28. Execution within entity orderedItems.add(128)
29. Execution within entity purchaseAgent.buy(“ 100€ lego“);
30. Execution within entity purchaseAgent.buy(“ 100€ lego“); java.net.ConnectException: connection refused
31. Execution within entity purchaseAgent.buy(“ 100€ lego“); java.net.ConnectException: connection refused event = new PurchaseRequestFailed(128) orderedItems.remove(128)
32. Execution within entity throw new IllegalStateException(“Order agent down”) Q?
33. How ESE make this easy? • Handles lifecycle of the entity • Entity instantiation • Initial replay of events • Initialization of entity • Handles request dispatching • Handles version of the entity and events • Every event increments version of the entity by one • Handles retries in case of errors • By default, every request is tried 5 times before it fails • Handles exceptions in all of the handling • Exceptions due to optimistic locks, execution of side effects, JSON serialization, …
34. The two building blocks Runtime class • Extends SomeEventSourcingRuntime • Singleton per entity type • Determines • request dispatch • entity lifecycle • event storage • request retries Entity class • Extends SomeEntity • One instance per id: String • Handles requests • Stores events • Builds state • Performs side effects
35. Entity initialization • If entity is not in memory, instantiate it, pass it dependencies • Replay all its events • (optimization possible with snapshots) • Invoke appropriate execution method * The things in italics is what you need to code
36. Dispatch support WishlistEntity extends AsyncEntity SyncEntity ProxiedSyncEntity<Wishlist> Request representation class VerifyParent implements Request<Parent> { String verificationToken(); } interface WishList { Parent verifyParent( String token); } Client interface parent = wishlists .execute(id, new VerifyParent( token)) .get() parent = wishlists .execute(id, new VerifyParent( token)) .get() parent = wishlists .execute(id) .verifyParent(token) Entity interface CompletableFuture<RS> execute( Request<RS> request) RS execute( Request<RS> request) WishList requestHandler() Runtime extends AsyncEventSourcingRuntime <WishlistEntity> SyncEventSourcingRuntime <WishlistEntity> ProxiedSyncEventSourcingRun- time<WishlistEntity,Wishlist>
37. Dispatch support • Utility classes for matching Request types to methods • Until pattern matching in Java 11… requestHandler = SyncRequestHandler.withDefaultFallback() .on(AddItem.class, this::addItem) .on(AdjustItemAmount.class, this::adjustItem) .on(RemoveItem.class, this::removeItem) .on(StartCheckout.class, this::startCheckout) .build(); requestHandler.handle(request); private OrderTotal addItem(AddItem request) { …
38. Persisting Events • persistAndUpdate(Event) • Store into • Memory (for tests) • JDBC • …
39. Events • Event consists of • Event type • Entity id • Entity version • Event timestamp • Payload attributes • Conveniently created with Immutables • https://immutables.github.io • Events are your API now. Evolve them carefully. @Value.Immutable public interface ParentRegisteredEvent extends GiftEvent { String id(); String name(); String email(); static Builder builder(EventSourcedEntity entity) { return new Builder().from(entity); } class Builder extends ImmutableParentRegisteredEvent.Builder implements FromEntity<Builder> { } }
40. Update state updateState(Event event) Similar helper methods for matching event types private TypeSwitch eventHandler = TypeSwitch.builder() .on(ParentRegisteredEvent.class, this::onParentRegistered) .on(ParentVerifiedEvent.class, this::onParentVerified) .build(); private Map<String, Parent> unverifiedParents = new HashMap<>(); void updateState(Event event) { eventHandler.executeMatching(event); } private void onParentRegistered(ParentRegisteredEvent event) { unverifiedParents.put(event.id(), Parent.of(ParentId.of(event.id()), event.name(), event.email())); }
41. Side effects You can perform side effects in your execute method, after persistAndUpdate finishes. performPostInvokeActions(List<Event> events, Throwable t) Fire and forget request-independent actions - no compensation possible here.
42. How to start <dependency> <groupId>io.github.goodees</groupId> <artifactId>ese-core</artifactId> <version>0.1</version> </dependency> <dependency> <groupId>org.immutables</groupId> <artifactId>value</artifactId> <version>2.5.6</version> <scope>provided</scope> </dependency> https://github.com/goodees/ese-example-gifts https://goodees.github.io/goodees/ese/apidocs/
43. Roadmap to 1.0 • Website • Full example • More convenient configuration of runtime • Support for building projections
44. Recap • The state is built from facts – past events • Every entity in ESE is a stateful object that performs business logic • Runtime is singleton that handles dispatch and configuration of environment • ESE gives you simple API for all of that Q?
45. Buzzword bingo score? Aggregate root Bounded Context Sharding CQRS Asynchronous Actor model Microservices Resilient Kafka Cloud native Scalable Schemaless Reactive Message passing Event store Event storming Replication Domain Driven Process manager Distributed High availability Lock-free CQS Eventual Consistency Stream processing
46. Thank you