Cost 20% more than RDS, delivers 5x and 3x performance improvement for MySQL and Postgres respectively
- Cloud Optimised, or AWS Optimised
- Drop in replacement for MySQL and Aurora (The drivers will work)
- Automatically grows upto 128TB from 10GB, no need to worry about size
- Can have upto 15 replicas, in sub-ms replica lag
- HA Native, Failover is instant
HA
- 6 Copies of data across 3 AZ
- Self healing in peer-to-peer replication
- One master instance takes the reads + 15 RR
- When primary instances fail, Aurora failovers to one of the reader instances
- Reader instances are replicated in async
- Continuos backup to s3
Aurora DB Cluster
- Writer endpoint is a dns name, when the master fails the dns points to the right master after failover
- Autoscaling can e setup for Read replicas, Aurora has reader endpoint also.
Aurora Backups and Restore
- PITR latest 5mins
- Retention perios 1-35 days, no disabling
Aurora Backtracking
- In Place restore, unlike backup
Aurora DB cloning
- Copy on write protocol
- Testing with prod data
- Old data:
Encryption
- Using KMS
- If master is not encyrpted, then rr can’t be encrypted
- To encrypt an unencrypt db, you’ll need to backup and restore with encryption
Failover
- Assign a priority tier on a read replica
- If two has the priority, auroro promotes the largest one.
- If both have same cap, then random
- RDS can be moved to Auroro