How do you handle exceptions, and where is the database connectivity?

Exception handling in a streaming job should be proactive and fault‑tolerant. First, wrap the core processing logic in a try/except block. Log the exception with context (e.g., record ID, timestamp) to CloudWatch or a centralized log store. For transient errors (network timeouts, temporary DB unavailability), implement an exponential back‑off retry strategy with a maximum retry count. For unrecoverable errors, route the problematic record to a dead‑letter queue (e.g., SQS or Kinesis DLQ) so that the job can continue processing other records.

Database connectivity is typically managed through a connection pool. In Glue, you can use the built‑in JDBC connector and configure the pool size via the --extra-jars and --conf options. Store credentials securely in AWS Secrets Manager and fetch them at job start. The connection pool ensures that connections are reused across micro‑batches, reducing latency and avoiding connection exhaustion. Always close the pool gracefully in a finally block or rely on Glue’s automatic cleanup.