Test building with shipwright

[olevski]

We have several options here:

• we can use the dev cluster to test
• we can simply build the images the same way the script in shipwright does it - just not publish it and we will not be able to run it (this is simplest)

Either way the main problem is that the way that pack builds images is very different than how Shipwright does it. Because Shipwright does not need a running docker daemon whereas pack does need it.

We have seen that especially image extensions are problematic in this way.

[sgaist]

There's a third option, renku-data-service's tests now have the option to deploy Shipwright.

If we want this to be fully self contained, we would need to also have Harbor deployed and some API calls to create the appropriate projects. As for accessing that instance from within the tests, the work done on SwissDataScienceCenter/renku-data-services#1264 and SwissDataScienceCenter/renku#4438 should all that is required.

[olevski]

@sgaist yes we can do that.

But we can also get by with the following:

• install the snb lifecycle executable
• run it so that it creates the oci image as layers somewhere locally
• podman (which is included in github ubuntu runners) can then just run the image when we point ti to the file
• then we can test

This will help because:

• we dont have to run docker in docker in the github runner anymore
• we can replicate the building that shipwright does