Flyte enables production-grade orchestration for machine learning workflows and data processing created to accelerate local workflows to production.

Flyte

Hi; question regarding SQLAlchemyTask — I was able to connect to postgres without issue, however I tried to connect to a MSSQL database using `pyodbc`, however it appears this is not available on the generic image that is user for those kinds of tasks. This makes sense given that this also requires a custom driver from MS to be installed to be useful.

Is it possible to swap out the docker image for `SQLAlchemyTask`? If so, how should I extend or build a new docker image for use with MSSQL?

you can provide your own image - <https://github.com/flyteorg/flytekit/blob/a6f9991952f6f95a2de99cf187e0778f3abff4e8/plugins/flytekit-sqlalchemy/flytekitplugins/sqlalchemy/task.py#L88|https://github.com/flyteorg/flytekit/blob/a6f9991952f6f95a2de99cf187e0778f3abff4e8[…]/plugins/flytekit-sqlalchemy/flytekitplugins/sqlalchemy/task.py>

Also I think we should just simply improve the entire sqlalchemy plugin, Flyte's fast registration system makes the plugin even simpler (the existing plugin was build on some legacy interface) - not wrong, just can be simplified.
cc <@U01J90KBSU9>

<@UNZB4NW3S> could you elaborate more on how fast registration will make the plugin simpler?

You don’t need the template passing that we do today. We always have the code now when we fast register and simply installing the plugin with alchemy can solve the problem. Also we can do
Similar to Spark where we can set the base image in imagespec

<@UNZB4NW3S> busy resurrecting a thread here but relevant to past conversation:

I’m trying to override `container_image` on the existing `SQLAlchemyTask` to use the image specified for the rest of the workflow, however I am getting an error like this:

```couldn't parse image reference "{{.images.default.fqn}}:{{.images.default.tag}}": invalid reference format```
The errors above are returned from Kubernetes so it looks like the `SQLAlchemyTask` is not resolving the templates. Any idea how I can get `SQLAlchemyTask` to run with the image passed in with `--image` without hard coding it?

My config follows:

```SQLAlchemyTask(
    "example.hello_mssql.sql_task",
    query_template="""
        SELECT @@version version
    """,
    container_image="{{.images.default.fqn}}:{{.images.default.tag}}",
    output_schema_type=DataSchema,
    task_config=SQLAlchemyConfig(
        ...
    ),
    secret_requests=[*secrets.values()],
)```

placeholder string in the image isn't supported in sqlalchemytask. <https://github.com/flyteorg/flytekit/blob/f721eef474bb04ec2df581a690fb703e7c2b7956/flytekit/core/python_customized_container_task.py#L163|this> needs to be wrapped in `get_registerable_container_image`<https://github.com/flyteorg/flytekit/blob/f721eef474bb04ec2df581a690fb703e7c2b7956/flytekit/core/python_auto_container.py#L263| method> for that to work. not sure what the implications are.

<@U05EFL3CVGQ> would you mind creating an issue? [flyte-core]

:star: Create a new Flyte Core Feature issue: <https://github.com/flyteorg/flyte/issues/new?assignees=&amp;labels=enhancement%2Cuntriaged&amp;template=feature_request.yaml&amp;title=%5BCore+feature%5D+>

<https://github.com/flyteorg/flyte/issues/4854>