In my project, it takes about `7 minutes to load the models when running the application locally.
Is there any way to improve the startup time? For example, is there a configuration that enables parallel model importing?
Here is my configuration:
mgmtp.a12.dataservices.initialization.import.models.path=classpath:/models/business/,\
classpath:/models/views/,\
classpath:/models/prints/,\
classpath:/models/mda/
And the log from dataservices:
Hello @nguyen-sharp-drift ,
There’s a ticket A12-18209. Could you please take a look?
In short, if your models contain a large number of localized fields (used in Enumeration fields) , it significantly slows down the model initialization process.
This will (presumably) improve with 2025.06-ext4 (DS 38.3).
Hi @patrick-long-drift , @loi-risen-dale
I upgraded from 2025-06-ext2 to 2025-06-ext4, (dataservices ‘38.3.4’), and the model import task kept running for a long time (more than 15mins), I have to manually stop the app.
It stuck at the step: “2026-03-24 15:28:35,204 [main ][INFO ][.initialization.BusinessModelInitializer][u:] - Importing business models: classpath:/models/business/”
Do you have any idea how to solve it?
Hi,
Our project also experienced significant performance issues, specifically when updating models. We traced the problem to a missing index on the LocalizedFieldEntity, which was causing a large number of sequential scans during the update process.
As seen in the LocalizedFieldsJpaRepository, this entity is not queried by its primary key, but rather by a combination of model_name, field_name, original_value, and locale.
By adding a composite index for these columns to the localized_fields table, we were able to significantly improve performance.
<sql>
CREATE INDEX CONCURRENTLY idx_localized_fields_lookup
ON localized_fields (
model_name,
field_name,
original_value,
locale
);
</sql>
Perhaps this solution could help with your performance issues as well.
Hi,
i tried the fix but it did not help (very much)
Is there a bugfix planed for a 2025.6 ext release?
Thanks
There is a fix in the 38.3.5 data services.