Amazon Aurora로 안전하게 migration 하기
3 likes574 views
The document details a technical migration process from Amazon RDS to Amazon Aurora, highlighting various architectural considerations such as database scaling and read replicas. It includes steps for setting up Aurora, monitoring replication lag, and promoting clusters, as well as comparisons of RDS and Aurora configurations. Resources and references to AWS documentation are also provided for further guidance.
Amazon Aurora로 안전하게 migration 하기
- 2. HBSmith . • HBSmith Co-Founder, CTO, SW Developer • 2013 AWS • #Architecture • : & , AWS, Python, Apple, Youtube • Email: jesang.yoon@hbsmith.io • GitHub ID: yoonjs2
- 3. , ? • HBSmith Business Logic & . • AI Bot 24 365 . • . • NaverLabs, KTH, , Wanted • hello@hbsmith.io https://hbsmith.io
- 4. . • AWS • Amazon RDS Database • Amazon Aurora • • Amazon Aurora
- 5. Computing : Disk • CPU • HDD • CPU HDD • SSD Reference: Intel newsroom
- 6. Backend Scale-out : RDBMS • Scale-out • RDBMS ( ) • DB Scale-out • Cache Sharding Read Replica
- 7. RDBMS : • RDBMS • ACID • Atomicity: - • Consistency: - • Isolation: - • Durability: -source: Loginworks software
- 8. RDBMS Scale-out ? : Split brain • Sync 2 master node • Sync • DB • Sync (Consistency ) • RDBMS “ ” • Split Brain . source: Trifork blog
- 9. : Split Brain Syndrome • (腦梁梁) • , • . source: Mrs. Eplin's IB Psychology Class Blog
- 10. RDBMS Scale-out ? : • Split brain Master 1 • Master • Master 1 SPOF • Master Scale-up • Scale-up • Sharding…
- 11. NoSQL RDBMS ?: CAP • DBMS 3 2 . • Consistency: . • Availability: . • Partition Tolerance: . source: Wikipedia
- 12. NoSQL RDBMS ? • NoSQL: CAP C A P ! • Consistency RDBMS • Consistency ...
- 13. Database ? • Amazon Aurora AWS ( ) • • AWS Amazon Aurora • (AWS Documentation ...)
- 14. Amazon Aurora Multi-master or Serverless
- 15. Amazon Aurora : Sixshop • Builder ( : WIX) • 2016 AWS • Amazon RDS MySQL • 2 • 10 DAU • DB
- 16. • Downtime • DBMS . • .
- 17. • 1 : DB Snapshot Aurora . • 2 : Aurora Replica Cluster . • Downtime 2
- 18. AS-IS : Spec • Risk . • DBMS • Database Engine Version • Network • DB ( , , ) • DB • DB Parameter
- 19. AS-IS : DBMS Parameter • : RDS Web Console Parameter -> 2 DB -> Compare • : RDS MySQL Parameter 1 , Aurora MySQL 1 , 1 • : encoding, engine, timezone .
- 20. AS-IS : Table Schema • • Schema ( ) • Table ( ) • Table Row Count
- 21. AS-IS : DB • • ? : Sixshop 3~4 • / ? : Sixshop 1~2 • ? : Sixshop • Architecture ? :
- 22. AS-IS : • DB VPC • DB VPC • : DB VPC VPC Peering Connection
- 23. TO-BE: • DB SQL Engine , Instance Tier Aurora • : Aurora MySQL 5.6 . 5.5 RDS MySQL 5.6 • AS-IS Step-by-Step . • AWS Root AS-IS TO-BE Step-by-Step . • : Step Step
- 24. : Step 0 - Migration • DB VPC • DB VPC • : DB VPC VPC Peering Connection
- 25. : Step 1 - Aurora Read Replica , • Aurora Read Replica migration • DB Aurora Cluster slave • Aurora Cluster master ( )
- 26. : Step 1 - Aurora Read Replica , source: AWS Documentation
- 27. : Step 1 - Aurora Read Replica , 1. Source DB Snapshot . 2. 1) Snapshop Aurora Read Replica . (Cluster Instance ) 3. Source DB Read Replica Replication = Source DB Replica Replica . Aurora Read Replica
- 28. : Step 2 - Replica Lag • Replication Replica Lag • Replica Lag 0 • Aurora Query Replica Lag Permission • CloudWatch
- 29. : Step 2 - Replica Lag : https://docs.aws.amazon.com/AmazonRDS/latest/UserGuide/Aurora.Monitoring.html
- 30. : Step 3 - Replica Lag 0 WAS • Replica Lag 0 DB • DB Write Transaction • • DB Write Transaction • WAS ELB
- 31. : Step 3 - Replica Lag 0 WAS • DB Row Count • Row Count • • Row Count Step SELECT CONCAT( 'SELECT "', table_name, '" AS table_name, COUNT(*) AS exact_row_count FROM `', table_schema, '`.`', table_name, '` UNION ' ) FROM INFORMATION_SCHEMA.TABLES WHERE table_schema = '<my table schema>';
- 32. : Step 4 - Aurora Master (Promote) • • DB Promote DB Cluster role • : Promote DB Query ! (Binary Log )
- 33. : Step 5 - Aurora Connection • WAS DB Endpoint DB • DB , ELB • DB Rollback
- 34. • Sixshop 2 ( 11GB) • DB • • • • & ( ) • DBMS Parameter • Amazon Aurora . .