Importing adaptive models to another environment in Pega 7.3 and later
You can import trained Adaptive Decision Manager (ADM) models from your production environment to a simulation environment. Synchronizing both environments is useful when you want to run scenarios in your simulation environment and apply the most up-to-date models. Adaptive models in the production environment are constantly processing data and self-learning; by importing these models to your simulation environment, you ensure that the scenarios that you run yield relevant and accurate results.
By following this procedure, you replace existing ADM data in the target environment with the ADM data that you import from the source environment.
Before you begin, make sure that the target environment meets the following requirements:
- The Cassandra cluster is not set up as Active-Active across multiple data centers. The following procedure does not apply to Active-Active setups.
- The adaptive rule versions are in sync with the versions in the source environment.
To import the models, perform the following actions:
- In the source environment, export the pyADMFactory data set. This data set is a database table that contains all the adaptive model instances in your system. For more information, see Exporting data into a data set.
- Log on to the target environment and perform the remaining steps there.
- On the Services landing page, on the Adaptive Decision Manager tab, decommission all ADM nodes by selecting the appropriate action from the Action menu.
- Open the pyADMFactory data set and from the Actions menu, select Run.
- In the pop-up window, from the Operation list, select Truncate.
- If the target system has any models that report data in the following tables, prevent inaccurate reports by manually truncating the following tables:
- pr_data_dm_admmart_pred_fact (in version 7.3)
- pr_data_dm_admmart_pred (in version 7.3.1 and later)
- Connect to a Cassandra database on a Decision Data Store node. For more information, see Configuring the Decision Data Store service.
- Remove any ADM (response) data that may cause a conflict with the source data by using the following CQL commands:
<span class="rulename">drop keyspace adm_commitlog</span>
drop keyspace null_adm(if present)
<span class="rulename">drop keyspace adm</span> (if present)
- Import the pyADMFactory data set from the source environment. For more information, see Importing data into a data set.
- Recommission all ADM nodes. The first node that you recommission creates scoring models from the imported factory data. For more information, see Configuring the Decision Data Store service.
What to do next: When the status of all ADM nodes is NORMAL, access the Model Management landing page and verify that the model data matches the source data. For more information, see Model Management landing page.
Published January 29, 2018 — Updated December 13, 2018