Feature Idea : Support Read Replicas

まず pg-pool の constructor、connect、query API を確認し、次に現在のクライアント選択と接続管理の仕組みを調べます。サポートする read-replica の構成と読み取り/書き込みのルーティング動作を定義します。これには、呼び出し側が isWrite を渡すのか、クエリを解析するのかも含まれます。完了条件は、既存の単一クラスター利用を壊さずにマルチクラスターのプーリングが機能することです。

node-postgres handles connection pooling seemlessly with a single cluster
Although it doesn't have support for multi cluster connection management
This will be useful in production deployments having multiple read replicas with a single master cluster used with write operations

Idea :
Allow a list of read clusters to be specified with pg-pool
with pg-pool handling connection management with multi-clusters automatically

Implementation Approach :
pg-pool has a constructor accepting a list of endpoints for read replicas , primary cluster ( used with write )

Internally A MultiClient Pool Manager will maintain the list of clients with public methods for picking a client for read or write
Client Pick Strategy For Read :
MultiClientPool will pick a cluster which has an idle connection if available , else it will resolve to creating a new connection by picking the one with min active connections

pg-pool has connect , query methods used by end clients , clients can have an option to specify an operation as read or write
Ex : through a parameter isWrite ,
pg-pool passes the parameter to MultiClientPool to pick a cluster

Alternative Approach to read-write detection

1. Implement a parser to auto identify a query as read or write

For Projects
It simplifies scaling read heavy workloads through usage of multi clusters