He Cao

> * [x] [api: add device CRD in scheduling group #376](https://github.com/koordinator-sh/koordinator/pull/376) > * [x] [api: add device info into NodeMetric CRD #378](https://github.com/koordinator-sh/koordinator/pull/378) hello, I am confused which GPU-related information is...

> This is somewhat captured in the previous `if` block, but we don't compose the logic in an `if...else...` manner, so yes, it's possible the failed != 0, but running/completed...

Also, I am wondering whether the current `Scheduling` phase is actually needed. I think maybe we can combine `PreScheduling` and `Scheduling` to one phase. A `Scheduled` Pod is a Pod...

> I think it is better to merge `Scheduling` and `PreScheduling` together. We should try to avoid updating the status of the pod group in the scheduler. I suppose the...

> when the creationTime are equal, it usually implies the two pods belong to the same PodGroup @Huang-Wei Thanks for your reply. I think in most cases, this is true....

Aha, actually I am not sure. I agree with you on the point that it's very unlikely to happen.

Thanks for ur feedback! The garbage collection mechanism for tenant's cluster-scope resources is not perfect. We will improve this and please feel free to contribute it.

To address this, the tenant controller needs to delete all kinds of cluster-scoped resources when a tenant is deleted. Considering there are 20+ kinds of cluster-scoped resources, I think maybe...

Thanks for pointing out it! We will add this restriction in the documentation. And it would be greatly appreciated if you could help improve the documentation :)

Thanks for your feedback! We will add more images to explain the design principle. And it would be greatly appreciated if you could point out which images are missing :)