Design Distributed Transactions with Sagas

Data, Storage & ConsistencyHarddesign-distributed-transactions

Design Distributed Transactions with Sagas

Design a booking flow that spans payment, inventory, and notification services, where there is no shared transaction to roll back — only compensations you have to write yourself.

SagaCompensating TransactionsTransactional OutboxTwo-Phase Commit

Traffic & Capacity Estimates:

5k multi-service transactions/second · 6 participant services · 99.999% settlement correctness

Functional Requirements

•Execute a multi-service workflow as a sequence of local transactions, each publishing an event.
•On any step's failure, run compensating transactions for every completed prior step, in reverse.
•Persist saga state durably so an orchestrator crash resumes rather than abandons in-flight work.
•Surface a terminal state (committed, compensated, or stuck) for every saga to operators.

Non-Functional Requirements

•No lost or duplicated effects — every step is idempotent and every event is delivered at least once.
•A single slow participant must not block unrelated sagas.
•Stuck sagas must be detectable and manually resolvable; silent partial states are unacceptable.

Back-of-the-Envelope Math

5k sagas/s * 6 steps = 30k step executions/second across the fleet.
At a 2% failure rate, 100 compensation chains per second must run correctly under load.

Key Architectural Trade-offs

Orchestration (a central coordinator — visible, testable, a new service to own) vs choreography (services react to events — decoupled, and nobody can explain the flow).
Sagas trade atomicity for availability: intermediate states are visible to users, so 'reserved' and 'pending' become part of your domain model.
Two-phase commit gives real atomicity where all participants support it, at the cost of blocking locks and an availability ceiling set by the weakest node.

Click or drag a component onto the canvas, then connect the handles to draw the data flow.

3 nodes · 2 edges

Components · 35

Client & Edge4

Compute & Gateway7

Storage & Caching11

Messaging & Streaming6

Coordination & Ops5

Intelligence2

React Flow