On Demand Feature Group Issue

Hi Team,

Sorry for clubbing 2 issues in same ticket:

Issue 1 (Within Hopswork Instance)

While trying to create a new on_demand_feature_group using mysql connector in Jupyter Notebook from Hopswork Instance always ending up with an error like “Access denied for user ‘“user1”’@’’ (using password: YES)”

The same mysql server can be accessed by tools like dbeaver with the same username and password.

FYI, the user1 has been granted with any ip permission in mysql (GRANT ALL ON . TO ‘user1’@’%’)

JDBC Connection String in Hopswork Storage Connector: jdbc:mysql://<mysql_ip>:3306/
JDBC Connection Arguments: user=“user1”
JDBC Connection Arguments: password=""

Not sure why from public demo instance alone(Hopswork Version = 2.1) we are facing this issue.

Issue 2 (From External Instance using python library)

While creating the same on_demand_feature_group from local machine using python library facing an issue as “TypeError: register_on_demand_temporary_table() missing 1 required positional argument: ‘alias’”

Steps Followed from demo notebook:

After getting the connection to hopswork instance in conn variable:
fs = conn.get_feature_store()
mysql_conn = fs.get_storage_connector(“mysql_cnnector”) – We have a storage connector created as mentioned in the above Issue
employee_mysql = fs.create_on_demand_feature_group(name=“employee_mysql”,
query=“select * from employee”,
description=“On-demand feature group for employee data”,

In the document didn’t find any argument like alias for the feature group creation (or) in get_storage_connector

Please help us to provide some spotlight here.

FYI, hsfs[hive] version 2.1.5


Hi @Arumugaguru_M,

If you can provide a stacktrace of the exception it would be better. It’s hard to tell what’s going on. If the username/password is correct, one possibility is that the version of the MySQL connector we provide (MySQL version 8) is not compatible with the version of your MySQL server.

We currently do not support configuring (i.e. calling .save()) and reading (i.e. calling .read() and .show()) on-demand feature groups using the Python engine. Currently only PySpark is supported.
If you are using the Python engine, you can still use the on-demand feature groups to create a training dataset.

Let me know if that’s helpful


1 Like

Hi Fabio,

Thanks, I was able to store the data now in on-demand Feature Group, I guess there were 2 issues,

  1. As mentioned by you, with driver, we have upgraded our MySQL instance to 8 from 5.6
  2. In the Storage Connector I had given the username & password with double quotes, and it also caused an issue


1 Like