Apache Airflow version
Other Airflow 2 version (please specify below)
If "Other Airflow 2 version" selected, which one?
2.7.2
What happened?
We are using 8.7.1. version of airflow provider package and set deferrable to True when using SageMakerTransformOperator. We set define the model name with something like "project-dag_name-date-uuid", where uuid is to deduplicate model names upon retry. However, when the execution gets back from deferred state, the original uuid is not preserved and there would be a new uuid so the operator cannot find a model with the new model name and it would fail.
What you think should happen instead?
The original UUID should be preserved
How to reproduce
In the operator config, set "ModelName": f"{name}-{str(uuid4())[:8]}". The run the DAG.
Operating System
amazon Linux AMI
Versions of Apache Airflow Providers
8.7.1
Deployment
Amazon (AWS) MWAA
Deployment details
No response
Anything else?
No response
Are you willing to submit PR?
Code of Conduct
Apache Airflow version
Other Airflow 2 version (please specify below)
If "Other Airflow 2 version" selected, which one?
2.7.2
What happened?
We are using 8.7.1. version of airflow provider package and set deferrable to True when using SageMakerTransformOperator. We set define the model name with something like "project-dag_name-date-uuid", where uuid is to deduplicate model names upon retry. However, when the execution gets back from deferred state, the original uuid is not preserved and there would be a new uuid so the operator cannot find a model with the new model name and it would fail.
What you think should happen instead?
The original UUID should be preserved
How to reproduce
In the operator config, set "ModelName": f"{name}-{str(uuid4())[:8]}". The run the DAG.
Operating System
amazon Linux AMI
Versions of Apache Airflow Providers
8.7.1
Deployment
Amazon (AWS) MWAA
Deployment details
No response
Anything else?
No response
Are you willing to submit PR?
Code of Conduct