amazon-documentdb

Run the amazon-documentdb-tools compat tool against your MongoDB source first to find unsupported operators, then pre-create indexes with the index tool. Set up a DMS full-load-and-cdc task with TLS verify-full on the target endpoint, monitor CDC latency, and cut over when lag reaches zero.

Lambda must run in the same VPC as the DocumentDB cluster, which has no public endpoint. Add an inbound TCP 27017 rule on the DocumentDB security group sourced from Lambda's security group ID, and include tls=true, the RDS global CA bundle, replicaSet=rs0, and retryWrites=false in the connection string.

Support varies by engine version, so check the MongoDB API compatibility page before relying on them. If unsupported, use the materialized ancestor path pattern for hierarchy queries and targeted compound indexes on known attribute paths instead of wildcard indexes.

COLLSCAN in explain() output means no index serves the query. Create a compound index following the ESR rule (equality, sort, range fields in order), then re-run explain() to confirm IXSCAN and check BufferCacheHitRatio in CloudWatch to verify the index stays in memory.

Use in-place modify-db-cluster with --allow-major-version-upgrade and a target-family parameter group, or the near-zero-downtime path: clone the cluster, upgrade the clone, replicate changes via DMS CDC, then cut over. Versions cannot be skipped, so 3.6 requires stepping through 5.0 first.

Elastic Clusters lack transactions, change streams, and many aggregation operators compared to instance-based clusters. The skill steers workloads to serverless or provisioned instance-based clusters instead, reserving Elastic Clusters only for sharding needs that exceed the largest instance classes.