Inqueue PodGroups occupy queue resources even when pods cannot be scheduled

[zjj2wry]

What is the problem you're trying to solve

When a PodGroup is moved to Inqueue state, it immediately reserves resources from the queue's capacity. However, if the pods cannot be scheduled due to reasons like:

• Node affinity/anti-affinity constraints
• Insufficient resources per node (e.g., requesting 8 GPUs but max 4 GPUs/node)
• ...

The PodGroup will stay in Inqueue state indefinitely, blocking queue resources that could be used by other jobs.（Resource waste）

Describe the solution you'd like

Implement an Inqueue timeout mechanism with the following behavior:

1. Queue-level default timeout (e.g., spec.inqueueTimeout: 10m
2. Record timestamp when PodGroup enters Inqueue state
3. Timeout action: When timeout is reached and no pods have been successfully scheduled, transition PodGroup back to Pending state
4. Release reserved queue resources

Additional context

No response

Documentation Updates

• This feature requires design or user documentation changes.
• If documentation changes are required, I will ensure the relevant documents are updated and published to the Volcano official website (https://volcano.sh) via the volcano-sh/website repository.

[JesseStutler]

Good cactch, yes, this is a question many users have met, and it's also mentioned in this issue: #4617. One approach is to roll back when scheduling fails, and another is to use the timeout mechanism you mentioned to flush back the pending status.

[JesseStutler]

We can also plan to implement it in the v1.15 version, do you have time to implement it?