Data Initialization Issue in CAP Java Project with HANA Database on Deployment

[stefaniaZ]

Hello SAP Community,

I am facing an issue with my CAP project (Java-based) where the database gets initialized with unexpected data on every deployment, specifically when the structure of a table is modified (e.g., adding new fields). Here is the context:

1. I created a CAP project with 4 tables, 3 of which are related to each other. Initially, the id field was of type Integer (uppercase), but it was later changed to UUID (lowercase).
2. For local development, I use SQLite, while in the production environment, I use HANA as the database.
3. Initially, we used .csv files to populate the tables, but since they were overwriting the database data on every deployment, we removed them.
4. Despite removing the .csv files, the issue persists: the database gets initialized with old data on every deployment, but only when the structure of a table is modified (e.g., adding new fields).
5. We had some "dirty" data in the tables with the old Integer IDs, which we manually deleted from the HANA database. However, after a new deployment, the "dirty" data reappeared. These records seem to match the data that was originally in the .csv files.
6. I have checked all the files in the project, and I cannot find any .csv files or other sources for this data. Yet, the data still gets initialized somehow.

   Question:
   How can I prevent the database from being initialized with this unexpected data on every deployment when modifying the table structure? Is there a configuration or process in CAP (Java) or HANA that could be causing this behavior?

   Thank you for your help!

[stefaniaZ]

Update:

• I tried adding empty .csv files and deleted all the data from the tables.
• After adding a new field to a table and deploying, the data was cleared as expected.
• I then reinserted data into the tables, deleted the .csv files, and redeployed. This time, the data was not cleared.
• However, when I modified a table again and redeployed, the data was cleared.

  In the deployer logs, I found the following message:

  Inserted 0 records of the data source file "myCsv.csv" into the target table "MY_TABLE"

  It seems like the .csv file is being cached or retained somewhere, even though it has been deleted from the project

[valentincadart · Accepted solution]

Hi @stefaniaZ,

A distinction must be made between test data (mocks) and initial real data:
The .csv files created in the test/data folder are only deployed when the application is not in production mode. This folder can therefore be used to store example data for tests and demonstrations.
The .csv files created in the db/data folder, on the other hand, are deployed in all environments, including production. This folder is therefore suitable for initial configuration data.

Have you tried cleaning the build artifacts: /gen, .cds deploy. There is also the --auto-undeploy parameter in cds deploy which allows you to : “Tell HDI deployer to automatically undeploy deleted resources”.

[stefaniaZ · Accepted solution]

Thank you,

I tried running the command:

cds deploy --auto-undeploy

and in the prompt, I saw the following message:

successfully deployed to db.sqlite

After performing the deploy, the tables were cleared as expected.

Here is my deployment procedure for all environments:

cf login --sso

mbt build -t gen --mtar mta.mtar (this generates a folder within the project containing the mtar file, which is the one I deploy)

cf deploy gen/mta.mtar

I opened the generated mtar file, but I do not see any .csv files inside it. However, the database still gets cleared during the deployment. Could you help me understand why this is happening?

Thank you again for your support!

[valentincadart · Accepted solution]

cds deploy --to hana --auto-undeploy : https://cap.cloud.sap/docs/guides/databases-hana#cds-deploy-hana

Maybe you can also cleaning the generated files before building the project using this documentation: "Delete generated files and directories of the previous build from the CAP Java project" => https://cap.cloud.sap/docs/java/assets/cds-maven-plugin-site/clean-mojo.html

In the mtar file you opened, did you see anything under the differents /resources paths ?

Another idea: try cloning the project and building/deploying it from another space...

Hope this helps you move forward with the issue

[stefaniaZ · Accepted solution]

Thank you for your suggestions @valentincadart ,

I can’t run your commands directly from the command line in when it does them on my local and not on cloud foundry.

Thanks to your advice, I was able to partially resolve the issue by enriching the undeploy.json file with the following two entries:

"src/gen/**/*.hdbtabledata",
"src/gen/**/*.csv"

However, I wanted to return to the initial setup to replicate the problem from the beginning. My goal is to get back to the state where I can delete the data, repopulate the file, and verify that the data is no longer being cleared.

After removing the two entries from the undeploy.json file and redeploying, I now encounter the following error:

Inserted 0 records of the data source file "myFile.csv" into the target table "MYTABLE"

However, it is not true that the data already exists or is duplicated. Do you have any idea why this might be happening?

Thank you,

Stefania

[valentincadart · Accepted solution]

Hi Stefania,

I came across some documentation that confirms using undeploy.json is indeed the correct approach to avoid deleting data during a redeploy:

• CAP Troubleshooting – undeploy.json
• SAP HANA Cloud Developer Guide – HDI Delta Deployment
• SAP Community – Data Deleted on CDS Deploy

Regarding your attempt to revert to the original setup where the CSV deployment repopulates your table, I found this documentation on .hdbtabledata which might help clarify how the import is behaving. In particular, please check the parameters like no_data_import and delete_existing_foreign_data, which could be affecting whether data is imported or skipped.

• .hdbtabledata Parameters

To debug further, you can unzip the generated MTAR archive and inspect the contents under: x-db-deployer/src/gen/data/your_csv.hdbtabledata. This file should show how the CSV is being treated during deployment.

Let me know what you find in those parameters—especially if no_data_import is set to true or if there's anything unexpected.

[stefaniaZ · Accepted solution]

Hi.

I tried creating the .hdbtabledata file, but two files keep getting generated, and I believe the wrong one is being used. When I open the .tar file, I see an .hdbtabledata file under src/data, which is the correct one, and another under src/gen/data, which is the incorrect one. I believe the system is using the wrong file. This is very strange behavior.

[stefaniaZ · Accepted solution]

Hi @valentincadart ,

The undeploy solution works, but the issue with data not being inserted anymore still remains. We are not sure how to properly use the .hdbtabledata file; if we manage to get it working, I will update the thread. Thank you for your support.

[valentincadart · Accepted solution]

Hi @stefaniaZ,

I just saw this recent blog about data loss when deploying from CAP to HANA .

The problem looks similar, I hope this helps.

[stefaniaZ · Accepted solution]

Hi valentincadart,

thanks so much.

Stefania