Sequence of Loading Models

Model load engine processes all the models asynchronously in parallel.

For performance reasons, all in-memory objects refer to the product using its md5 checksum. When the rule models are processed, you have to make sure the corresponding product is present in the cache before the rule is processed. If the product is not found for the corresponding rule, model processing throws ModelProcessorException.

After the model processing is complete, MLE notifies Consul to update the model key - /config/application/notification/modelsProcessed - for respective tenant and document type. This will inform all the running OPE instances to copy updated memory objects.

Note: In case of model processing error - where few models are moved to failure bucket, refresh notification will not be sent out. Therefore, you will need to modify and re-publish them so that refresh notification is sent out for all OPE instances.