Feature: Add connection libraries to the superset instance

[Maleware]

Current Situation

if a user want's to have e.g. mssql as a database connector present in a superset instance, he'd need to either

1. Build a custom image using pip install pymssql
2. kubectl exec -- pip install pymssql
3. Overrite startup commands to execute pip install pymssql

a list of possible connector options can be found here

Proposal

As a platform provider we should enable a user to be flexible and on the other hand ensure that our images are as lean and secure as possible.

Option 1

Include common connectors in our image.

Upside

1. User has maintained and out of the box images that "just work" for their purposes
2. Stackable has full control over versioning and can ensure and guarantee that specific scenarios will work
3. Low hanging fruit ( up to decision on what include, that might be bigger than it seems )

Downside

1. We'd need to be responsible for upcoming bugs, CVE's as well as maintaining library versions.
2. Larger images and delivery of not needed software

Option 2

Allow users to specify python packages to install in crds

Upside

1. Users can specify the libraries they want to be installed
2. Lean and secure base images
3. No maintenance on Stackable end needed
4. Mechanism might be reusable for Airflow

Downside

1. Open up the installation to malicious software
2. Needs to be an Override as we can't take guarantee of user defined packages
3. Larger development efforts expected

[stefanigel]

I like the idea of making users' life easier. I like option 2 more than option 1, as it also maybe a license issue as there are many commercial db drivers in the world.
Adding db drivers / libs / connections may easily become a never ending story as every user might look for sth different.

[Maleware]

Me too, but we'd need to estimate effort and security. I like that this could be reused in python based products.