There might be a use case, we may need to copy/move the ADM learning from PRODUCTION environment to other lower environments to simulate the results OR Reproduce Some scenarios referring to ADM learning from Production environment.
The following steps should be sufficient to import
learning to a target system even if that system has previously had ADM models
itself. However, they assume that you want to completely replace any ADM data
on that target system with data from the source system, not merge the two. It
is also assumed that your Cassandra cluster is not set up for Active/Active,
across multiple datacenters.
In
the source system
1. Export the pyADMFactory dataset from the source
System.
In
the Target system
1. Ensure
the adaptive rules are present.
2. Decommission
all ADM nodes.
3. Navigate
to the pyADMFactory dataset and run
truncate operation.
4. If
the target system has any model reporting data in the following tables, these
should also be manually truncated to prevent misleading reports.
a. pr_data_dm_admmart_mdl_fact
b. pr_data_dm_admmart_pred_fact
c. pr_data_dm_admmart_pred
(this may not exist depending on the version of Pega)
5. Connect
to Cassandra on one of the DSS nodes
a. Remove
the ADM model’s response lifecycle event data
drop keyspace adm_commit_log.
b. Remove
the rest of the ADM Data
drop keyspace_null_adm
and/or drop keyspace adm - which ever present
6. Navigate
to the pyADMFactory data set and run
the import data using the data exported from the source system.
7. Decommission
ADM nodes (the first node to start will create scoring models from the imported
factory data).w
8. When
the ADM nodes have status NORMAL,
check the Adaptive Model Management landing page. Model data should now reflect
what was imported
