CraftConf Budapest 2018: Break your event chains! Complex event flows in distributed systems

@berndruecker @martinschimak => May 10th, 2018 => Budapest => CraftConf

Transcript

1. Break your event chains! Complex event flows in distributed systems. @berndruecker @martinschimak
2. Berlin, Germany bernd.ruecker@camunda.com @berndruecker Bernd Ruecker Co-founder and Developer Advocate of Camunda
3. Vienna, Austria martin.schimak@plexiti.com @martinschimak Martin Schimak Into and onto de|coding domain language with Plexiti
4. Distributed systems
6. Let‘s have a look what we can learn from DDD
7. Example process / Event Storming! Goods shipped Order placed Goods fetched Payment received Events are a concept easy to grasp for everybody in the room.
8. Events are … Payment received Payment Strong consistency It already happened. Events let us look into the past.
9. Events help to decouple services. Order summary Payment received Payment UI raises projected onto Payment Service Order Service
10. Events help with cross-cutting concerns. Payment received Inform customer listened to by Payment Service Notification Service
11. What about purely event-driven architectures? Order placed Payment Service Payment received Inventory Service Goods fetched Shipment Service Goods shipped When relying only on events, „intent“ is always formed inside the event consumer.
12. Autonomous services can make their own decisions. To change the order of two services, we touch 4 sticky notes… but wait, even three services! Shipment Service Goods shipped Order placed Inventory Service Goods fetched Payment Service Payment received
13. Our policies decide to act on facts – with commands Domain Event Policy Command listened to by triggers
14. Who owns the policy in a pure event choreography? Payment received Start Fetching Policy Fetch goods listened to by triggers Payment Service Inventory Service
15. Neither of both truly „owns“ that policy? Payment received Fetch goods listened to by triggers Payment Service Inventory Service Start Fetching Policy
16. Extracting the policy into a mediator service Payment received Fetch goods listened to by triggers Payment Service Inventory Service Order Service Start Fetching Policy
17. The order as the company‘s core business Place order Order placed Request payment Order Service Payment received Fetch goods Goods fetched Ship goods Goods shipped “When the order is placed…” “When the payment is received…” “When the goods are fetched…” “Then we are done!” Mark order as fulfilled Order fulfilled Payment Service Inventory Service Shipment Service
18. Double check: what happens when we charge later? Place order Order placed Request payment Order Service Payment received Fetch goods Goods fetched Ship goods Goods shipped “When the order is placed…” “When the payment is received…” “When the goods are fetched…” “Then we are done!” Mark order as fulfilled Order fulfilled Payment Service Inventory Service Shipment Service “When the order is placed…” “When the payment is received…” “When the goods are fetched…”
19. Multiple services need to collaborate Payment Shipping Order Inventory
20. Multiple services need to collaborate Payment Shipping Order Inventory https://www.infoworld.com/article/3254777/application-development/ 3-common-pitfalls-of-microservices-integrationand-how-to-avoid-them.html
21. Photo by Tookapic, available under Creative Commons CC0 1.0 license.
22. „There was an error while sending your boarding pass“
23. Check-in Web-UI Me Current situation
24. Check-in Barcode Generator Web-UI Me Output Mgmt Current situation
25. Check-in Barcode Generator Web-UI Me Output Mgmt Current situation – the bad part
26. Check-in Barcode Generator Web-UI Me Output Mgmt Current situation – the bad part
27. Check-in Barcode Generator Web-UI Me Output Mgmt Current situation – the bad part Stateful Retry
29. We are having some technical difficulties and cannot present you your boarding pass right away. But we do actively retry ourselves, so lean back, relax and we will send it on time. …I just made this up…
30. Check-in Barcode Generator Web-UI Me Output Mgmt Possible situation – much better!
31. Check-in Barcode Generator Web-UI Me Output Mgmt Possible situation – much better! Stateful Retry
32. Check-in Barcode Generator Web-UI Me Output Mgmt Stateful Retry Possible situation – much better! The failure never leaves this scope!
33. Persist thing (Entity, Document, Actor, …) State machine or workflow engine Typical concerns DIY = effort, accidental complexity Complex, proprietary, heavyweight, slow, don‘t scale, developer adverse Scheduling, Versioning, operating, visibility, scalability, … Handling State
34. Workflow engines, state machines It is relevant in modern architectures
35. CADENCE Silicon valley has recognized Workflow engines, state machines
36. CADENCE There are lightweight open source options Workflow engines, state machines
37. CADENCE also at scale Workflow engines, state machines
38. CADENCE for todays demo Workflow engines, state machines
39. Check-in Barcode Generator Web-UI Me Output Mgmt A state machine!
40. Check-in Barcode Generator Web-UI Me Output Mgmt A state machine!
41. Example Credit Card Payment REST also available in C#
42. Live hacking https://github.com/flowing/flowing-retail/
43. Payment Now you have a state machine! Credit Card REST
44. Distributed systems introduce complexity you have to tackle! Credit Card Payment REST
45. Distributed systems
46. It is impossible to differentiate certain failure scenarios. Independant of communication style! Service Provider Client
47. Distributed systems introduce complexity you have to tackle! Credit Card Payment REST
48. Distributed systems introduce complexity you have to tackle! Credit Card Payment REST Do it reliably
49. Distributed systems 2007
50. Distributed transactions using compensation * Compensation *aka Saga pattern
51. Live hacking https://github.com/flowing/flowing-retail/
52. Long running services Credit Card Payment Order
53. God services? Credit Card Payment A few smart god services tell anemic CRUD services what to do Sam Newmann Order
54. God services? Credit Card Payment Order A few smart god services tell anemic CRUD services what to do Sam Newmann
55. „By delegating our problems to our clients we force them into mitigation. They become god services.“ Credit card failed Retry with new credit card details Charge credit card listened to by triggers Payment Service Order Service
56. Just the first requirements for a payment service Request payment Payment received Payment requested Charge credit card Depending on account balance Withdraw amount from account Amount Withdrawn Credit card charged Credit card failed Update credit card details Account details Credit card details updated Whenever card details are updated After two weeks Depending on the amount Amount credited Mark payment received Mark payment canceled Payment canceled Credit amount to account Whenever payment is canceled Mark payment received Payment received
57. Payment as a „long-running“ service Request payment Payment received Payment canceled 1) creates value – serving a result its client is really interested in 2) assumes responsibility – instead of delegating troubles with internal steps to its client 3) may go asynchronous – if needed for composite execution Payment Service
58. @Saga class Payment { private var paymentAmount = 0F @StartSaga @SagaEventHandler(associationProperty = "paymentId") fun on(event: PaymentRequested) { } } e.g. Axon Messaging & Saga Management is cool, but Java works as well! :-) 1) Correlate domain event 2) Save process state 3) Evaluate business policy 4) Create and issue command
59. Credit amount back to account Withdraw amount from cust. account Charge amount by credit card e.g. with Camunda as message-driven workflow Charge amount by credit card Credit amount back to account Update credit card details Check customer account balance Payment requested Depending on account balance Charge credit card Withdraw amount from account Credit card charged Whenever credit card is charged Mark payment received Payment received Amount Withdrawn Depending on the amount Mark payment received Payment received Mark paym. partly covered Payment partly covered Whenever payment p. covered Charge credit card Credit card failed Whenever credit card failed Update credit card details Credit card details updated Whenever card details are updated Charge credit card After two weeks Mark payment canceled Payment canceled Whenever payment is canceled Credit amount to account Amount credited Every two days Remind customer Customer reminded Withdraw amount from cust. account
60. DDD? Long-running services respect bounded contexts! Order service Payment service Request payment Payment received Request payment Payment canceled
61. Integration of Axon Messaging with Camunda. Order service Payment service Request payment Payment received Payment canceled Request payment plexiti de|coding domain language @martinschimak https://github.com/plexiti/axon-camunda-poc
62. Flowing retail example: https://github.com/flowing/flowing-retail/ Using Java + Spring Boot or C# REST or Kafka Camunda or Zeebe
63. Zeebe.io
64. bernd.ruecker@camunda.com @berndruecker http://bernd-ruecker.com/ martin.schimak@plexiti.com @martinschimak http://plexiti.com/ https://github.com/plexiti/axon-camunda-poc/ http://github.com/flowing/flowing-retail/ https://www.infoq.com/articles/events- workflow-automation Contact: Contact: Examples: https://www.infoworld.com/article/3254777/ application-development/ 3-common-pitfalls-of-microservices- integrationand-how-to-avoid-them.html With thoughts from http://flowing.io @berndruecker | @martinschimak