Migration from GCP sql (postgres17.7) to AWS rds (postgres17.7) using AWS dms

To improve your PostgreSQL migration from Google Cloud SQL to Amazon RDS using AWS DMS, you need to optimize several parameters and consider a few key adjustments.

For your current "migrate only" (full load) approach:

1. The slow migration speed you're experiencing (only a small fraction of data transferred after 18 hours) suggests performance issues. For large tables like yours (8515 MB, 7697 MB, etc.), consider:

• Dividing large tables into separate tasks using selection rules
• Monitoring disk and CPU usage on the target side

2. For LOB handling, your settings (LOBChunkSize of 64KB and Maximum LOB size of 0) are appropriate for full mode if you have large objects, as this allows unlimited LOB sizes.

For when you move to "migrate and replicate" (full load + CDC):

Source database (Google Cloud SQL) parameters:

• You've already enabled "cloudsql.logical_decoding" and "cloudsql.enable_pglogical" which is good
• Ensure binary logging is properly configured

Target database (Amazon RDS) parameters:

• Increase max_logical_replication_workers
• Increase max_sync_workers_per_subscription
• These parameters help speed up initial data syncing during replication

Additional considerations:

1. Database extensions are not replicated as part of pglogical replication. You'll need to manually create them in your target RDS instance.
2. For minimal downtime, after your full load completes, you can set up CDC to capture ongoing changes while your application still uses the source database.
3. When using CDC, you'll need to provide a native start point or choose "Immediately" in the data migration settings.
4. Monitor replication status and CDC latency before performing the final cutover.
5. Remember that AWS DMS creates a serverless environment for the migration process, using pg_dump/pg_restore for schema migration and a publisher/subscriber model for ongoing replication.

For your specific case with large tables and public endpoints, ensure your security groups and network settings allow proper connectivity between the DMS replication instance and both databases.
Sources
Migrating data from PostgreSQL databases with homogeneous data migrations in AWS DMS - AWS Database Migration Service
Migrate Google Cloud SQL for PostgreSQL to Amazon RDS and Amazon Aurora using pglogical | AWS Database Blog
Migrate a self-managed MySQL database to Amazon Aurora MySQL using AWS DMS homogeneous data migrations | AWS Database Blog
Step-by-step PostgreSQL database to Amazon RDS migration walkthrough - Database Migration Guide

Your current DMS task is running very slowly because it’s using Full LOB mode with Max LOB size = 0, which forces DMS to stream each large object row-by-row without batching. That’s safe for truncation but extremely inefficient for multi‑GB tables. For initial migration, switch to Limited LOB mode with a realistic Max LOB size (for example, 1–2 MB if your largest LOBs fit) so DMS can batch and parallelize reads. You can then identify any outliers with larger columns and handle them separately if needed. Also make sure your replication instance has enough vCPU and memory (use at least a `dms.r5. —Taz