Migrate ROCm-RAG Weaviate data

2025-09-26

2 min read time

Applies to Linux

You can migrate data between different storage locations in a cluster through Weaviate, the AI-native vector database used by ROCm-RAG, or through local raw data migration.

Cluster migration (backup and restore)

You can move entire clusters across different storage backends with Weaviate’s built-in backup and restore functionality.

Raw data migration (export and import)

You can specifically move raw data objects by exporting collections to a JSON file and importing them into another machine.

For example, you can export one collection from the Weaviate database and save it to /rag-workspace/rocm-rag/backup/rocm_rag_backup_data.json on the source machine with the Weaviate instance running:

python /rag-workspace/rocm-rag/rocm_rag/data/weaviate_export_data.py

Then you can load the .json file and import it into the target machine’s running Weaviate instance:

python /rag-workspace/rocm-rag/rocm_rag/data/weaviate_import_data.py

This approach is helpful when you only need to migrate the dataset without transferring the cluster configuration or storage state.