A saga is easy to draw. It's brutally hard to make correct.
Idempotency both ways, out-of-order compensations, failures you can't tell apart, rollbacks that fail too.
New bonus piece:👇https://t.co/QGCFC8K0Wa
Stop solving microservice consistency in your application code.
AI writes much of that layer now, and it mostly works. But "mostly" isn't "correct" — and the people closest to the code usually can't tell the difference.
Series finale 👇
https://t.co/xAf2fyFo8h
Sagaと2PCはオペレーションによって使い分けるものでどちらか一方の身を常に使用するものではないという趣旨の記事。Oracle AI DatabaseはSagaも2PCも可能という意味ではこの主張に添えるかと思います。もちろんScalarDBと組み合わせて使っていただいてもかまいません。
https://t.co/M8QX8f38mJ
"Saga or 2PC?" is the wrong question.
The right one is per operation: what does this piece of work actually need?
A real system needs both — and most of the case against 2PC is really a case against one 30-year-old implementation.
New post 👇
https://t.co/rGlL0saVNB
"Just use Saga, avoid 2PC."
The default advice for consistency across services. I think both halves are too simple.
Inside one DB, "all or nothing" is free. Across services, it isn't.
New post on why the framing gets both sides wrong 👇
https://t.co/lyKkmrsYoF
"Don't reach for microservices too early — start with a monolith."
For most of the last decade, that was the right call.
I think AI is quietly changing it.
I wrote up why, as Part 1 of a new series, Architecture in the AI Era. 👇
https://t.co/vYn75x6MXA