Proxy Advisor Workflow

When This Applies

User mentions: “should I use RDS Proxy”, “too many connections”, “connection pooling RDS”, “RDS Proxy pinning”, “Lambda database connections”, “RDS Proxy cost”, “RDS Proxy for PostgreSQL/MySQL”. Not for Aurora — Aurora has its own proxy considerations (Aurora Serverless interactions, Global Database).

Tasks

1. Gather Instance Metrics

Pull connection data to assess whether proxy is needed.

Constraints:

• You MUST ask for the DB instance identifier and region upfront
• You MUST run aws rds describe-db-instances to get engine, instance class, and vCPU count
• You MUST pull CloudWatch metrics for the last 7 days from the AWS/RDS namespace:
  • DatabaseConnections (Average, Maximum; for p99 pass it via --extended-statistics p99, not --statistics) — current connection count on the DB
  • CPUUtilization (correlate connection spikes with CPU)
• If the user is considering migration away from an existing RDS Proxy, you MUST also pull these proxy metrics (AWS/RDS namespace, dimension ProxyName):
  • ClientConnections — frontend connections to the proxy
  • DatabaseConnections — backend connections the proxy holds (distinct from the DB-side metric)
  • DatabaseConnectionsCurrentlySessionPinned — the canonical pinning diagnostic; high values mean multiplexing is defeated
• You MUST determine max_connections for the instance class. The RDS default is roughly LEAST({DBInstanceClassMemory/9531392}, 5000). Check the parameter group for overrides.

Example CLI to pull the DB-side connection metric:

For proxy-side metrics, swap the dimension to Name=ProxyName,Value=<proxy-name>. For p99, drop --statistics and pass --extended-statistics p99 instead.

• You MUST calculate connection utilization: peak_connections / max_connections × 100%

2. Assess Proxy Need

Constraints:

• You MUST categorize the result using these thresholds:
  • 🔴 Recommended: peak connections > 80% of max_connections, OR Lambda/serverless callers present
  • 🟡 Consider: peak connections 50–80% of max_connections, OR frequent connection churn (spiky workload)
  • 🟢 Not recommended: peak connections < 50% and stable connection patterns
• You MUST ask if the application uses Lambda or other serverless compute, because connection churn from cold starts is the primary RDS Proxy use case
• You MUST check whether the application already uses client-side connection pooling (PgBouncer, ProxySQL, HikariCP, etc.) — if it does and connections are healthy, proxy may add latency without benefit
• You MUST NOT recommend proxy on utilization metrics alone — low utilization with heavy Lambda churn still benefits from proxy

3. Check for Pinning Risks

Connect to the database or ask the user to run diagnostic queries to identify SQL patterns that cause proxy pinning.

Constraints:

• You MUST explain what pinning is: the proxy pins a frontend connection to a specific backend connection, preventing multiplexing — the proxy’s core benefit
• For MySQL/MariaDB, check for patterns from proxy-pinning-mysql.md
• For PostgreSQL, check for patterns from proxy-pinning-postgresql.md
• You MUST categorize pinning risk as High / Medium / Low
• If pinning risk is High, you MUST warn that proxy benefit will be significantly reduced and quantify where possible (e.g., “if >30% of connections pin, the multiplexing advantage is largely negated”)
• For PostgreSQL: you MUST call out SET search_path as the most common high-pinning pattern (Django, Rails ORM defaults pin every connection). Recommend moving it to the proxy init query or parameter group default.
• For MySQL: you MUST call out server-side prepared statements as the most common high-pinning pattern (JDBC default). Recommend useServerPrepStmts=false or client-side prepared statements.
• When the user asks how to measure pinning in production, you MUST direct them to the CloudWatch metrics DatabaseConnectionsCurrentlySessionPinned and ClientConnections (AWS/RDS namespace, dimension ProxyName) and the RDS Proxy pinning log events. You MUST NOT fabricate a pinning-rate percentage when no live metrics have been pulled.

4. Estimate Cost

Constraints:

• You MUST calculate proxy cost based on vCPU count of the target instance. RDS Proxy pricing is $0.015 per vCPU per hour in us-east-1 (verify for other regions; pricing varies slightly).
• Monthly cost = vCPUs × $0.015 × 730 hours
• You MUST present cost alongside the benefit (connection multiplexing value, failover improvement ≈ 66% faster than direct)
• You MUST note that proxy adds ~5 ms latency per connection establishment — for latency-sensitive workloads with short connection lifetimes, this can matter

5. Special Cases

• Already using PgBouncer in transaction mode: RDS Proxy’s multiplexing benefit is marginal. PgBouncer in transaction mode is actually more aggressive (no pinning on SET). RDS Proxy adds value for managed infrastructure, IAM auth, and Multi-AZ failover handling — but it’s a “Consider”, not a strong “Recommend”.
• TLS to the backend: RDS Proxy enforces TLS between proxy and database by default — a security advantage over self-managed PgBouncer, which may not enforce backend TLS.
• Advisory locks (PostgreSQL) or GET_LOCK() (MySQL): High pinning — connection is pinned for the lock’s entire hold time. Recommend replacing with application-level locking (Redis, DynamoDB).
• LISTEN/NOTIFY (PostgreSQL): Pins the backend. Replace with SQS/SNS/EventBridge if using proxy.

6. Present Recommendation

Constraints:

• You MUST present a clear verdict: Recommended / Consider / Not Recommended
• You MUST include: current connection utilization, pinning risk level, estimated monthly cost, and key tradeoffs
• You MUST NOT create an RDS Proxy — this workflow is advisory only
• You SHOULD reference proxy-pinning-mysql.md or proxy-pinning-postgresql.md for the engine-specific pinning taxonomy

Troubleshooting

• No CloudWatch data: Instance may be newly created. Ask user for expected connection patterns instead.
• max_connections unclear: Check parameter group with aws rds describe-db-parameters. If not custom, use the engine default formula based on instance memory.
• User unsure about pinning: Provide the diagnostic queries from the pinning references and ask them to run and paste results.
• Multi-AZ with proxy: Proxy handles failover automatically — additional benefit worth mentioning in the recommendation.
• IAM auth required: RDS Proxy is the supported path for IAM database authentication on RDS MySQL/PostgreSQL — mention as a benefit when applicable.

References

• proxy-pinning-mysql.md — MySQL/MariaDB pinning patterns and diagnostics
• proxy-pinning-postgresql.md — PostgreSQL pinning patterns, search_path gotcha, extended query protocol notes