Group Replication原理解析与实践经验 徐春阳 MGR原理 MGR vs SEMI-SYNC MRG实践经验 • 支持多节点并发执行事务。如何保证从各个节点并发执行 的事务在每个节点以相同的顺序被应用/执行.------paxos协 议的功能。 • 自动地事务冲突检测。节点之间不能”同时”操作主键相同 记录,如果冲突，只有一个节点成功。------冲突检测。 mysql_parse! mysql_execute_comma nd! trans_commit_stmt! MYSQL_BIN_LOG::com mit! group_replication_trans_before_commit! waitTicket! ���������� ���������� ���������� ������paxos� ��������� ���������� primary���MGR���� ����������������! MGR�������������� ����������! MRG原理 MGR vs Semi-Sync MGR实践经验 Group_replication_bootstrap_group参数须谨慎 DDL操作注意事项 当心secondary节点变成可写节点 规避大事务 Secondary��������! ������Single

和/或其附属公司。保留所有权利。 3 MySQL 复制 MySQL 半同步 复制 Windows/ Solaris/ Clusterware 群集或 Oracle VM Group Replication MySQL Cluster MySQL 高可用性解决方案 4/16/2017 9 9 . 9 9 9 % 版权所有 2015，Oracle 和/或其附属公司。保留所有权利。 5 数据冗余存储 冗余应用服务器 数据路由 4/16/2017 版权所有 2015，Oracle 和/或其附属公司。保留所有权利。 7 • MySQL Replication • Shared Storage • Group Replication • MySQL Cluster 从数据库 • 会话线程：处理来自应用程序的查询 － 将 数据写入主数据库，将关联事件写入二进制 日志 • 转储线程：读取二进制日志中的事件，然后 – 不需要代理 •可选的代理方式能提供高级 主机/OS 监视 4/16/2017 版权所有 2015，Oracle 和/或其附属公司。保留所有权利。 12 Enterprise Replication Monitor •自动搜索复制拓扑 •主/从性能监视 •复制指导 •有关最佳做法的复制建议 “我每天使用 MySQL Enterprise Monitor 监视和跟踪 MySQL

& access control • RBAC: admin, developer, guest • AD/LDAP integration • Policy based image replication • Web UI ( 中文 and English) • Audit and logs • Restful API for integration • Lightweight and easy Replication Service 15 Remote Harbor instance Image Replication between Registry Instances 16 Project Images Policy Image Project Images Initial replication Image Incremental Incremental replication (including image deletion) Image Replication Use Case(1) • Image distribution for large cluster • Load balancing 17 Master – Slave Docker Client push Docker host Docker

Join us! • Submit a talk Postgres Conference China 2016 中国用户大会 “We need to scale” “We need replication” Postgres Conference China 2016 中国用户大会 What are your requirements? Postgres Conference China Can new PostgreSQL 9.6 features help? Postgres Conference China 2016 中国用户大会 PostgreSQL 9.6 Replication • New option in 9.6: synchronous_commit = remote_apply – remote_write: write, but no fsync – on: database and reading from a replica that the data might not yet have be applied.. • Synchronous replication (on) may reduce write transaction throughput. Using remote_apply potentially more so • If it

本⽂概述了Kubernetes集群中所需的各种组件。 Master组件 Master组件提供K8s集群的控制⾯板。Master对集群进⾏全局决策（例如调度），以及检测和响应集群事件（例如：当 replication controller所设置的 replicas 不够时，启动⼀个新的Pod）。 Master可在集群中的任意节点上运⾏。然⽽，简单起⻅，设置脚本通常在同⼀个VM上启动所有Master组件，并且不会 个单独的进程，但为了降低复杂性，它们都被编译成独⽴的⼆进制⽂件并运⾏在⼀个进程中。 这些控制器包括： Node Controller：当节点挂掉时，负责响应。 Replication Controller：负责维护系统中每个replication controller对象具有正确数量的Pod。 Endpoints Controller：填充Endpoint对象（即：连接Service＆Pod）。 Service 译者按：FQDN是fully qualified domain name的缩写，即：完全限定域名 并⾮所有对象都在Namespace中 ⼤多数Kubernetes资源（例如：Pod、Service、Replication Controllers等）都在某些Namespace中。但Namespace资 源本身并不在Namespace中。低级资源（例如： nodes 和persistentVolumes）也不在任何Namespace中。事件是⼀个

available - Refreshing replication account of demoted cluster - Synchronizing transaction backlog at 127.0.0.1:4442 - Updating metadata - Updating topology -- Changing replication source of 127.0.0.1:3331 1:3331 to 127.0.0.1:4442 -- Changing replication source of 127.0.0.1:3333 to 127.0.0.1:4442 -- Changing replication source of 127.0.0.1:3332 to 127.0.0.1:4442 - Acquiring locks in replicaset instances replica clusters -- Changing replication source of 127.0.0.1:5552 to 127.0.0.1:4442 -- Changing replication source of 127.0.0.1:5553 to 127.0.0.1:4442 -- Changing replication source of 127.0.0.1:5551 to

& access control – RBAC: admin, developer, guest – AD/LDAP integration • Policy based image replication • Vulnerability Scanning • Notary • Web UI • Audit and logs • Restful API for integration Replication Job Services Notary client Remote Harbor Instance Notary Registry V2 Vulnerability Scanning Admin Service Harbor users and partners (selected) 12 Image replication (synchronization) (synchronization) 13 Project Images Policy Image Project Images Initial replication Image incremental replication (including image deletion) Agenda 1 Container Image Basics 2 Project Harbor Introduction

and sources and targets for real-time reporting and analytics using Log based CDC and Data Replication. What is HVR? 4 Key Benefits of Using HVR for your Business 提升业务洞察力 关键业务连续性 提高效率 降低风险 5 • Incompatible with new big-data technologies (EMR, Spark, etc..) Added value from HVR • Rich replication (next slide) • Multiple SCD models (slides after that...) • Good “dove-tailing” with subsequent must be implemented to compact changes into a presentation layer. – Latency is higher – Basic replication (not rich) means higher TCO Added value from HVR • Timekey Compare: row-by-row compare of source

Point of Failure Asynchronous Replication A = 1 B = 2 A = 1 Master Slave Asynchronous Replication A = 1 B = 2 A = 1 B = ? Master Slave Synchronous Replication A = 1 B = 2 A = 1 B = 2 Master Master Slave Synchronous Replication A = 1 B = 2 A = 1 B = 2 Master Slave Serving? Slave -> Master? Huge Amounts of Data Master Slave Master Slave Master Slave •Open-source NewSQL database isolation/node failure tiggers election •Log replication •Leader handles client requests, append logs •Replicate logs to followers •Constraints •Quorum: log replication, leader election •Up to date: leader

iteblog_keyspace WITH replication = {'class': 'SimpleStrategy', 'replica tion_factor': 1}; cqlsh> 上面命令创建了名为 iteblog_keyspace 的 keyspace；并且采用 SimpleStrategy 进行副本复制，因为我们这个测试集群只有单个节点，所以这里设置的副本因子（replication factor）为 KEYSPACE iteblog_keyspace 时，按上键盘上的 Tab 键，cqlsh 会自动给我们不全到 CREATE KEYSPACE iteblog_keyspace WITH replication = {'class': '，这时候我们再按 Tab 键，会显示出支持的所有副本复制策略。具体大家可以去试试。 创建完 keyspace 之后，我们可以使用 DESCRIBE KEYSPACE DESCRIBE KEYSPACE iteblog_keyspace; CREATE KEYSPACE iteblog_keyspace WITH replication = {'class': 'SimpleStrategy', 'replication_fa ctor': '1'} AND durable_writes = true; 现在我们可以使用 USE 命令来切换到这个 keyspace