-
Notifications
You must be signed in to change notification settings - Fork 96
Description
The current structure of the examples and demo won't scale as we continue to add examples:
The opaque names gpu-test1.yaml, gpu-test2.yaml, etc. are hard to reason about without the diagram. The examples should have reasonably descriptive names that don't require some sort of separate index to figure out what they are. Supporting commentary for each example should live closer to the rest of that example and separate from other examples instead of all together in the README.
As we add examples, more mock devices are required to fit all of the examples at once. With the direction we're going, the driver will eventually advertise a couple dozen "GPUs" on each node by default, which isn't realistic and makes reading the ResourceSlices burdensome. Adding more nodes to the default kind cluster config isn't much better since the driver should run on any cluster. IMO the default number of devices should at least fit the largest example, but should not need to satisfy running every example all at once. We should update the demo to show a limited number of representative examples that can be installed with the default instead of all of the examples.
Metadata
Metadata
Assignees
Labels
Type
Projects
Status