2.2 Distributed Transaction Solution: TCC

1. What Is TCC?

A distributed transaction solution that guarantees eventual consistency.

2. TCC Process

• TCC: Try, Confirm, Cancel.
• Divide the transaction into two phases:
  • In Phase 1, execute Try to perform business checks and reserve resources.
  • Phase 2 depends on the result of Phase 1.
    • If Phase 1 succeeds, execute Confirm to perform business confirmation.
    • If Phase 1 fails, execute Cancel (perform an operation opposite to Try, similar to a manual rollback).
  • 

3. 2PC vs TCC

• It is very similar to 2PC; both have two phases.
  • 2PC is a two-phase approach at the database layer, while TCC is a two-phase approach at the application layer.
  • Each Try, Confirm, and Cancel in TCC runs in a separate transaction, while 2PC operates within one transaction.
• TCC solves several shortcomings of 2PC:
  • Performance problem: Phase 1 only checks and reserves resources; it does not need to lock them until Phase 2.
  • Single-point problem: after Phase 1 is complete, if the business activity manager goes down in Phase 2, there are still other business activity managers.
  • Consistency problem: after Phase 1 is complete, if any participant goes down in Phase 2 and does not return an ACK, keep retrying until it succeeds (because Phase 1 has already reserved the resources and can guarantee success).

4. Problems with TCC

• Highly intrusive: depends on the business side to cooperate by providing Try, Confirm, and Cancel interfaces.

5. Use Cases for TCC

• Suitable for business results that require strong consistency, high real-time performance, and whose distributed transactions can be rolled back, such as the three core services in Internet finance companies: trading, payments, and accounting.
• Multiple services use multiple data sources, and the data sources do not have to be DBs.

6. Implementations of TCC

6.1. Seata TCC Mode

• Does not depend on transaction support from the underlying data resources:
  • Phase-one prepare behavior: call custom prepare logic.
  • Phase-two commit behavior: call custom commit logic.
  • Phase-two rollback behavior: call custom rollback logic.

6.2. Seata AT Mode

• Based on relational databases that support local ACID transactions.
• Java applications access the database through JDBC.
• In Seata’s AT mode, we do not need to write Phase 2 at all; it is entirely implemented by Seata itself.

7. References

• If Someone Asks You About Distributed Transactions Again, Send Them This Article - Juejin
• Transactions - What Is the Difference Between TCC and 2PC? - SegmentFault
• What Is the Difference Between TCC and 2PC? - Zhihu
• Seata TCC Mode
• Seata AT Mode