RDS Operator

Install and Deploy

Option 1: kubectl way

1. kubectl apply -f examples/operator.yaml ( provide aws credentials in here )
2. kubectl apply -f examples/auroraMySQLCluster.yaml

Option 2: Helm chart

---

Example/Samples

• Can be found under examples/ dir

DBCluster vs DBInstance

• Use DBCluster and DBInstance (Both) WHEN creating anything with aurora ( so aurora-mysql or aurora-postgresql )
• Why use DBCluster? Because of how AWS SDK works.
• Their docs for DBCluster state the following

    //
    // Valid Values: aurora (for MySQL 5.6-compatible Aurora), aurora-mysql (for
    // MySQL 5.7-compatible Aurora), and aurora-postgresql
    //
    // Engine is a required field
    Engine *string `type:"string" required:"true"`
  

• So the only valid values for DBClusters are aurora-mysql and aurora-postgresql
• Once DBCluster is created, you must attach instances to that cluster, thats where DBInstance comes into the picture.
• How to attach a DBInstance to a DBCluster? You have to provide DBClusterIdentifier within DBInstance.
• WHEN to use standalone DBInstance? -- WHENever you want to create anything thats not aurora.

Features

• Create ( All RDS Databases, Subnet groups )
  • DBCluster: spec.createClusterSpec ✅
    • credentialsFromSecret: spec.credentialsFrom ✅
    • externalNameService: spec.serviceName ✅
    • initDBJob spec.initDBJob ❌
  • DBInstance: spec.createInstanceSpec ✅
    • credentialsFromSecret: spec.credentialsFrom ❌
    • externalNameService: spec.serviceName ✅ ( primary db instance )
    • initDBJob spec.initDBJob ❌ ( if attached to DBCluster, this wont run at all. )
• Delete ( All RDS Databases, Subnet groups ) ✅
  • DBCluster: spec.deleteClusterSpec ✅
  • DBInstance: spec.deleteInstanceSpec ✅
• Restore from rds snapshot ( All RDS Databases )
  • DBCluster: spec.createClusterFromSnapshot ✅
  • DBInstance: spec.createInstanceFromSnapshot ✅

TODO

• Reconcile AWS resources periodically ❌ ( Recreate DB incase they get deleted )
  • Need some sort of trigger that the cloud resource got deleted.
  • For the time being, use k8s objects as events to trigger reconcile instead of periodic sync?
• Centralize (create/delete/restore) into rdsLib and make use of existing interface funcs! ✅
• Add docs
• Make Secret resource conditional for both DBCluster(:white_check_mark:) and DBInstance ( some folks might want to pass credentials like username and password from a k8s secret so there is no need to deploy another secret with the same information )
• Add support for initDB Job in dbCluster so that a user can have their db imported from an existing image ( DBCluster & DBInstance )
• Make delete optional and get rid of deleteSpec to enforce snapshotting before deleting rds.
• Update ( COMING SOON ) ❌
• CreateReadReplica WHEN deployed from scratch and or from snapshot ( COMING SOON ) ❌