Design a data pipeline to ingest and process data from multiple sources (e.g., S3, Kinesis) to Redshift using Spark.

Section 1 — The Context (The 'Why')
Multi-source pipelines ingest from heterogeneous systems (S3, Kinesis, databases, APIs) into a unified store. The primary challenge is coordinating different latencies (batch vs. stream), schemas, and SLAs. A shared-nothing design breaks when a slow S3 scan blocks Kinesis consumers; backpressure and isolation between sources are essential.

Section 2 — The Diagram

[S3] ------> [Glue Crawler] ----> [Bronze Lake]
[Kinesis] -> [Lambda/Flink]  -->  |
[RDS] -----> [DMS/CDC] ---------> |
[API] ------> [Kinesis/EventBridge] -> |
                    |
                    v
            [Unity Catalog / Schema Reg]

Section 3 — Component Logic
Each source has a dedicated ingestion path. S3 uses Glue crawlers or Lambda triggers; Kinesis consumers write with different prefixes. CDC streams RDS changes. A central schema registry enforces evolution. Idempotency via primary keys and merge semantics. Partitioning strategies vary by source: date for logs, (tenant, date) for multi-tenant. TTL and lifecycle policies tier data to Glacier. Fan-out: bronze feeds multiple silver consumers without re-ingestion. In production, monitor consumer lag, checkpoint success rate, and sink write latency as primary SLOs. Partitioning strategies should align with query patterns; bucketing within partitions mitigates join skew. TTL policies on raw and intermediate data control storage cost while preserving replay capability for debugging and backfill. Data skew mitigation via salting or secondary hashing prevents single partitions from becoming bottlenecks. Exactly-once semantics require transactional commits at the sink; at-least-once delivery demands idempotent write logic to avoid duplicates. Fan-out patterns allow one source topic to feed multiple downstream consumers without re-ingestion. Backpressure handling ensures that slow processors do not cause unbounded buffer growth; Kafka consumer lag is a key metric. Schema evolution should follow additive-only rules where possible to avoid breaking consumer compatibility. The CAP trade-off should be documented per component: analytics typically favors AP, while financial reconciliation requires CP. Blast radius from component failure is bounded by replication and checkpointing; design for graceful degradation during partial outages. Cost optimization: use Spot instances for batch workloads and tier cold data to lower storage classes. Dead-letter queues preserve failed records for replay rather than dropping them.

Section 4 — The Trade-offs (The 'Senior' part)