Configure MongoDB Atlas connection pooling and enable retryable writes for a serverless or short-lived application

Verified steps

1. Use the official MongoDB driver for your language and construct the connection string with the +srv DNS seedlist URI format: mongodb+srv://<user>:<password>@<cluster>.mongodb.net/<db>?retryWrites=true&w=majority
2. Create the MongoClient instance once at application startup (singleton pattern) — do not create a new client per request; the driver maintains a connection pool internally and reuses connections across requests
3. Set pool size parameters in the connection string or client options: maxPoolSize (default 100) controls the maximum number of connections; set it based on expected concurrency and your Atlas tier's connection limit; serverless/FaaS environments should set maxPoolSize=1 or use the Atlas Data API to avoid pool exhaustion
4. Enable retryable writes (retryWrites=true, on by default since driver 4.x) so the driver automatically retries a write once on a network error or primary failover — this covers the most common transient failure modes transparently
5. Set serverSelectionTimeoutMS (default 30,000 ms) and connectTimeoutMS to values appropriate for your latency budget; in Lambda or Cloud Run where cold starts occur, the first connection attempt must complete within these timeouts
6. Use connection string options readPreference=secondaryPreferred for read-heavy workloads to distribute reads to replica set secondaries; pair with maxStalenessSeconds to bound how stale secondary data can be