Migrating to Aurora, Checklist for Success

在过去的几周中,我们向您展示了如何成功地从本地MySQL安装迁移到AWS Aurora。接下来要做的是如何成功地确保您的AWS Aurora集群按照您期望的方式执行和运行。
虽然Aurora的无手操作方法确保了敏捷实践保持敏捷,但也存在权衡和随后的增长难题。
本次网络研讨会将讨论如何在使用AWS Aurora时保持灵活性并完全控制数据。

展开查看详情

1.Migrating to Aurora, Checklist for Success Jervin Real Percona, Sept 2018 1 / 19

2. Before You Deploy https://www.percona.com/resources/technical- presentations/migrating-aws-aurora-monitoring-aws-aurora- pmm-percona-technical 2 / 19

3. Goals Complete control of our data. We must not be operationally constrained by most common operations. Predictable performance. 3 / 19

4.General AWS Considerations 4 / 19

5. Costs Considerations Instance pricing, upfront vs on demand IOPs/million Backup storage Backtrack Cross region replication Storage rate $/GB/mo 5 / 19

6.#1 Reduce the Unknowns 6 / 19

7. Review Queries Regularly Bad queries are one of the top offenders of downtime. Before and after code changes. Keep historical query performance data. 7 / 19

8. Review Queries Regularly Performance Insights Query Analytics from Percona Monitoring and Management https://aws.amazon.com/blogs/database/analyze-amazon-aurora-mysql- workloads-with-performance-insights/ Dashboard shown is custom dashboard. https://pmmdemo.percona.com/ 8 / 19

9.#2 Optimize for Binary Log Replication 9 / 19

10. Optimize for Binary Log Replication Independent backup node. Reporting or analytics. Cross region redundancy. 10 / 19

11. Optimize for Binary Log Replication binlog_format = ROW Multi-threaded replication. Performance metrics from PMM/Performance Insights helps a lot. 11 / 19

12.#3 Design for Flexibility 12 / 19

13. Design for Flexibility Disable lab mode. Separate parameter groups per cluster. Keep your dataset size in check. Disable auto minor version upgrade Use the right tools. 13 / 19

14. Use the right tools. Aurora cannot remove a misbehaving read-replica from the load balanced reader-endpoint. A bad query is a bad query, it can spoil your day. Sometimes, Aurora failover is not fast enough - or a cluster restart is not acceptable. 14 / 19

15.Wrapping Up 15 / 19

16. To Recap Reduce the Unknowns Optimize for Binary Log Replication Design for Flexibility 16 / 19

17.Questions! 17 / 19

18.18 / 19

19. Percona is Hiring! Consultants, Support Engineers and DBAs MySQL MongoDB PostgreSQL and more ... https://www.percona.com/about-percona/careers 19 / 19