## Group Replication原理解析与实践经验 徐春阳 ## QCon 全球软件开发大会 10⽉17-19⽇上海·宝华万豪酒店  扫码锁定席位 ## 九 折即将结束 团购还享更多优惠，折扣有效期至9月17日 扫描右方二维码即可查看大会信息及购票 用户线程执行事务（sql）流程  group_replication_trans_before_commit 被唤醒后 将有关事务的日志写入到本地通道。后面的所有任务，例如，将这个事务日志利用paxos协议进行全局一致性广播，验证是否跟其他事务冲突等全部 CONTENTS ## MRG原理 MGR vs Semi-Sync MGR实践经验 ## TABLE OF CONTENTS ## 当心secondary节点变成可写节点 ## Group_replication_bootstrap_group参数须谨慎 规避大事务 DDL操作注意事项 ## Secondary节点变成可写节点 正常情况下，Single primary 模式下，secondary节点只读

20 MoveTables ..... 20 Identifying Candidate Tables ..... 20 Query Rewriting ..... 21 Replication Graph ..... 25 Shard ..... 25 Shard Naming ..... 25 Resharding ..... 26 Tablet ..... 26 a CNCF project .54 Scalability Philosophy .54 Small instances ..... 54 Durability through replication ..... 54 Consistency model ..... 54 Multi-cell ..... 55 Supported Databases ..... 56 MySQL 60 Advanced usage ..... 61 Undocumented features ..... 61 Known limitations ..... 61 Replication ..... 61 Semi-Sync ..... 61 Database Schema Considerations ..... 62 Point In Time Recovery

Partitioner Types 12 About the Random Partitioner 12 About Ordered Partitioners 13 About Replication in Cassandra 13 About Replica Placement Strategy 14 SimpleStrategy 14 NetworkTopologyStrategy Partitioner Settings 25 Snitch Settings 26 Configuring the PropertyFileSnitch 26 Choosing Keyspace Replication Options 27 Installing and Initializing a Cassandra Cluster 27 Installing Cassandra Using the Cluster 95 Calculating Tokens For the New Nodes 96 Adding Nodes to a Cluster 96 Changing the Replication Factor 97 Replacing a Dead Node 97 Backing Up and Restoring Data 98 Taking a Snapshot 98

Authentication ..... 639 20.4. Resource Consumption ..... 646 20.5. Write Ahead Log ..... 655 20.6. Replication ..... 666 20.7. Query Planning ..... 673 20.8. Error Reporting and Logging ..... 680 20.9 and Point-in-Time Recovery (PITR) ..... 802 27. High Availability, Load Balancing, and Replication ..... 813 27.1. Comparison of Different Solutions ..... 813 27.2. Log-Shipping Standby 901 30.5. WAL Configuration ..... 902 30.6. WAL Internals ..... 905 31. Logical Replication ..... 907 31.1. Publication ..... 907 31.2. Subscription ..... 908 31.3. Row Filters

364 14.4.6. Increase checkpoint_segments .....364 14.4.7. Disable WAL archival and streaming replication .....364 14.4.8. Run ANALYZE Afterwards .....364 14.4.9. Some Notes About pg_dump .....365 Settings .....432 18.5.2. Checkpoints.....435 18.5.3. Archiving .....436 18.5.4. Streaming Replication.....437 18.5.5. Standby Servers .....438 18.6. Query Planning .....438 18.6.1. Planner Method 24.4.2. Other data migration methods.....533 25. High Availability, Load Balancing, and Replication.....534 25.1. Comparison of different solutions.....534 25.2. Log-Shipping Standby Servers

release notes Highlights: • Bug Fixes and improvements as usual • ARTEMIS-4136 Mirror sync replication Mirror now has an option to set sync=true. Blocking operations from clients will wait a round Mirror support improving capabilities on AMQP Mirror for Disaster Recovery • Journal Retention • Replication integrated with Zookeeper • Connection Routers • Concurrency configuration for core bridges. queues • AMQP Server Connectivity "Basic" SecurityManager implementation that supports replication • Consumer window size support for individual STOMP clients • Improved JDBC connection management

20 MoveTables ..... 20 Identifying Candidate Tables ..... 20 Query Rewriting ..... 20 Replication Graph ..... 25 Shard ..... 25 Shard Naming ..... 25 Resharding ..... 26 Tablet ..... 26 project ..... 90 Scalability Philosophy ..... 90 Small instances ..... 90 Durability through replication ..... 91 Consistency model ..... 91 Multi-cell ..... 92 Supported Databases ..... 92 MySQL ... 102 3. Push-based metrics system ..... 102 Monitoring with Kubernetes ..... 102 Replication ..... 102 Semi-Sync ..... 102 Database Schema Considerations ..... 103 Point In Time

[Image](/uploads/documents/4/8/4/c/484c9d62390ce6f7d8cf5e435c4c269f/p5_1.jpg) • MySQL Replication 数据冗余存储 • Shared Storage • Group Replication • MySQL Cluster ## MySQL 复制工作流 二进制日志 会话 转储 中继日志  • Group Replication • MySQL Cluster # 主动 接管 共享 存储 被动 版权所有 2015, Oracle 和 / ## 适用于 MySQL 的 Oracle VM 模板 • 预安装且预配置

14.4.6. Increase checkpoint_segments ..... 339 14.4.7. Disable WAL archival and streaming replication ..... 339 14.4.8. Run ANALYZE Afterwards ..... 340 14.4.9. Some Notes About pg_dump ..... 340 Settings ..... 403 18.5.2. Checkpoints ..... 406 18.5.3. Archiving ..... 407 18.5.4. Streaming Replication ..... 407 18.5.5. Standby Servers ..... 408 18.6. Query Planning ..... 409 18.6.1. Planner 24.4.2. Other data migration methods ..... 498 25. High Availability, Load Balancing, and Replication ..... 499 25.1. Comparison of different solutions ..... 499 25.2. Log-Shipping Standby

379 14.4.6. Increase checkpoint_segments .....379 14.4.7. Disable WAL Archival and Streaming Replication .....379 14.4.8. Run ANALYZE Afterwards .....379 14.4.9. Some Notes About pg_dump .....380 5.1. Settings ..... 449 18.5.2. Checkpoints ..... 453 18.5.3. Archiving ..... 453 18.6. Replication ..... 454 18.6.1. Master Server ..... 454 18.6.2. Standby Servers ..... 456 18.7. Query Planning archive_command Scripts .....555 24.3.6. Caveats .....555 25. High Availability, Load Balancing, and Replication .....557 25.1. Comparison of Different Solutions .....557 25.2. Log-Shipping Standby Servers