Migration to Storage v2
Make sure to follow the Repository section from the DSE Controller Configuration page.
Use this procedure to migrate existing trial results from KAI DC Builder version 1.x to version 2.0. It assumes you are using the recommended deployment method with the ${REPODIR}/aidc/.env configuration file for Docker Compose.
-
Edit the
.envto make sure the STORAGE variable points to a path to a new storage location. If you're using the standard template for theenv.2.0, it would look like this: -
In the current shell session, initialize a variable with a path to the existing
v1storage location – the same way it was used to run KAI DC Builder version 1.x. For example: -
Copy the existing storage data to the new storage location. If you were using multiple DSE server names for different set of results with v1, it is a good moment to decide if you want to migrate all of them or only results from some servers. Choose subdirectories below accordingly. The example below copies results from all servers via a
*wildcard. -
Clean the existing virtual environment to make sure all dependencies are reinstalled for version 2.0.
-
Deploy KAI DC Builder v2.0 from a Docker Compose setup with a Jupyter Notebook server. First, only start the DSE server.
cd ${REPODIR}/aidc docker compose -f compose.nb.yml create docker compose -f compose.nb.yml start dse docker compose -f compose.nb.yml logs -f dseObserve the logs to ensure the server starts correctly – you should see a message below. Interrupt the log streaming with
Control-C. -
Pull the required Python modules from the DSE and install Jupyter notebooks. Start the Jupyter server. This step may take several minutes.
cd ${REPODIR} && sudo -E make pull-modules notebooks cd ${REPODIR}/aidc docker compose -f compose.nb.yml start jupyter docker compose -f compose.nb.yml logs -f jupyterWait until the package installation is complete and you see the message below. Interrupt the log streaming with
Control-C. -
Open the Jupyter Notebook server in your browser, depending on the hostname of system with the DSE. Open the
migrate_storage.ipynbnotebook and follow the instructions there to complete the migration process. -
After the migration is complete, delete the
v1subfolder under the storage location to free up space. This will still keep the original data intact in the previous storage location defined by theSTORAGE_V1variable.
Data Duplicates
The migration process creates new IDs for each trial result. Therefore, if running the migration multiple times, you may see duplicate results.