Distributed Systems

1. What Is Saga - A distributed transaction solution that guarantees eventual consistency. 2. Saga Process - There are multiple transaction participants, and each participant has two pieces of logic: a forward operation and a reverse operation. - Divide the transaction into two phases.

1. What Is 3PC - A distributed transaction solution that guarantees strong consistency and is an improved version of 2PC. 2. 3PC Process - Split the first phase of 2PC into two steps, so the whole transaction process has three phases: CanCommit, PreCommit, and DoCommit.

1. What Are Two Phases - Divide the entire transaction process into two phases: preparation phase; commit phase or rollback phase. 2. Two-Phase Implementations 2.1. Database Layer - Two-Phase Implementation: 2PC. 2.2. Application Layer - Two-Phase Implementation: TCC.

1. What Is a Local Message Table - Use two transactions, placing an order and then adding points, as an example. The ordering service is Service A and the points service is Service B. 1. Service A executes ordering logic. 2. After Service A successfully places the order, it sends a message to MQ. 3. Service B consumes the message and processes its local transaction.

1. What Is 2PC - 2PC: Two-Phase Commit. - Divide the transaction into two phases. - In the first phase, the transaction manager sends prepare requests to all databases. - If all respond ok, execute commit in the second phase; if any responds fail, execute rollback in the second phase.

1. What Is TCC - 2PC is a two-phase approach at the database layer, while TCC is a two-phase approach at the application layer. - TCC: Try: attempt to execute the transaction; Confirm: confirm execution of the transaction; Cancel: cancel execution of the transaction. - In essence, it is also a two-phase transaction.

1. What It Is - 1. The producer finishes executing its local transaction and sends a message to MQ. 2. MQ sends the message to the consumer. 3. The consumer consumes the message and executes its local transaction; if successful it ack's, and if it fails it nack's and requeues the message. 4. The consumer can actively call the producer's interface to query message status.

1. What Are RocketMQ Transaction Messages - Traditional local message tables depend on a message table in the database. - RocketMQ transactions encapsulate the local-message-table approach by moving the local message table into MQ, solving the atomicity problem between Producer-side message sending and local transaction execution.

1. What Are Distributed Consensus Algorithms - More precisely, they are consensus algorithms: making all nodes agree on something. 2. Why Consistency Problems Occur 2.1. Concurrent Client Requests - For example, in a Leader-Follower scenario: one Client, three Nodes A, B, and C. The Client asks A to write x as 1. If A considers x to be 1, then B and C must also consider the value to be 1.

1. Basic Paxos 1.1. What Is Basic Paxos - Abbreviated as Paxos. - A distributed consensus algorithm invented by Lamport and the foundation of Raft and ZAB. 1.2. Basic Paxos Algorithm Process 1.2.1. Roles - client: request initiator; not important here. - proposer: proposal proposer, similar to a coordinator.