Import dashboard with overwriting all related objects (charts and datasets)

[rad-flu-alu]

Hi everyone.

I have two instances of Superset. On the first instance I have a dashboard which I exported to a zip file. As we know, the file contains the definition of the dashboard and all of the related objects, i.e. charts, datasets and database. I imported the dashboard into the second instance of Superset - the dashboard, charts and datasets were created. On the second instance I changed the dashboard and the charts.
Now I want to export the dashboard from the second instance and import it into the first instance of Superset. When I use the overwrite option the dashboard is changed, as expected, but the charts are not modified and the dashboard and charts are not consistent.

How can I overwrite all of the objects during the import of the dashboard?

Regards,
Rafał

[dosubot[bot]]

Hello @rad-flu-alu! I'm here to help you with any bugs, questions, or contributions you have. Let's tackle this dashboard import issue together.

To overwrite all related objects, including charts and datasets, when importing a dashboard in Superset, you should use the ImportAssetsCommand with the overwrite=True parameter. This ensures that all existing objects are replaced with the new data being imported. Here's a general approach you can follow:

1. Export the Dashboard: Export the dashboard from the second instance of Superset, which includes all related objects like charts and datasets.

2. Prepare the Import: When importing the dashboard into the first instance, ensure that the import process is configured to overwrite existing objects. This is typically handled by the ImportAssetsCommand in Superset, which uses the overwrite=True parameter for each type of asset (databases, datasets, charts, and dashboards).

3. Run the Import Command: Execute the import command with the necessary configurations. The command should look something like this:

   from superset.commands.importers.v1.assets import ImportAssetsCommand
   import yaml
   
   # Define your configurations
   metadata_config = {...}
   database_config = {...}
   dataset_config = {...}
   chart_config = {...}
   dashboard_config = {...}
   
   # Prepare the contents for import
   contents = {
       "metadata.yaml": yaml.safe_dump(metadata_config),
       "databases/imported_database.yaml": yaml.safe_dump(database_config),
       "datasets/imported_dataset.yaml": yaml.safe_dump(dataset_config),
       "charts/imported_chart.yaml": yaml.safe_dump(chart_config),
       "dashboards/imported_dashboard.yaml": yaml.safe_dump(dashboard_config),
   }
   
   # Create and run the import command
   command = ImportAssetsCommand(contents)
   command.run()

This approach ensures that all related objects, including charts and datasets, are overwritten during the import process ^[1][2].

To continue talking to Dosu, mention @dosu.

[illaaa]

Hello @AparnaSharma1501 !
Could you share some detail of how you did that, please?
The dosubot suggested to use /import/ endpoint, but if I understood it right, you were able to import using /dashboard/import/.
And what's the crucial fix that helped you to import the dashboard with underlying objects overwriting?
I would be super appreciate for your help!

[katmolony]

+1

Hello @AparnaSharma1501 ! Could you share some detail of how you did that, please? The dosubot suggested to use /import/ endpoint, but if I understood it right, you were able to import using /dashboard/import/. And what's the crucial fix that helped you to import the dashboard with underlying objects overwriting? I would be super appreciate for your help!

[illaaa]

For those who come looking for a solution — I wasn't able to import a dashboard along with all its dependent objects using the /dashboard/import/ endpoint. Also, according to the API documentation, there is no such endpoint as /import/; instead, there is /assets/import/.

I couldn’t get it to work via /assets/import/, but maybe I missed something.

As an alternative solution, I import the objects sequentially: database → dataset → chart → dashboard, using their corresponding endpoints (/database/import/, /dataset/import/, /chart/import/, /dashboard/import/). One important detail here is that each import package must contain only its respective lower-level objects and nothing else. Additionally, the type attribute in the required metadata.yaml file must differ for each package.

As an example, here's the dictionary I used: (import_package_name, required_folders, import_package_type):

@katmolony FYI

[coderAnunay]

@illaaa Thanks for the interesting insight, would it possible to share some examples how you have structured the objects in the zip file? I am facing a similar issue as yours related to importing charts

[illaaa]

@illaaa Thanks for the interesting insight, would it possible to share some examples how you have structured the objects in the zip file? I am facing a similar issue as yours related to importing charts

The objects are structured the same as they are when you export each separate entity, i.e. dashboard package should look like that:

|—archive.zip
|—|—any_folder_name
|—|—|—metadata.yaml
|—|—|—dashboards
|—|—|—|—my_dashboard_1.yaml
|—|—|—charts
|—|—|—|—my_chart_1.yaml
|—|—|—|—my_chart_2.yaml
|—|—|—databases
|—|—|—|—my_database_1.yaml
|—|—|—datasets
|—|—|—|—my_database_1
|—|—|—|—|—my_dataset_1.yaml

For charts the structure is the same but without dashboards folder, and etc. for other import packages. And keep in mind the "type" attribute in metadata.yaml, it's different for each different "package type".

Also, if you want to import a few entities of the same type (i.e. 5 dashboards), you can put everything in one package and import it once.