E2E for Volcano scheduler in different sharding mode

[qi-min]

Summary

The existing Volcano scheduler has been extended to support shard-based pod scheduling. To validate this enhancement and prevent regressions, we need to expand the current E2E test suite to cover pod scheduling behavior under sharding configurations.

Scenarios

• No Sharding

Verify that when no sharding is configured, Pods are scheduled to the nodes within the cluster according to the default scheduling and scoring policies.

• Hard Sharding Mode

Verify that when hard sharding is enabled, Pods are scheduled strictly to nodes within the assigned shard, and no nodes outside the shard are selected. e2e need cover allocate, preempt, reclaim, backfill actions

• Soft Sharding Mode – Preferred In-Shard Scheduling

Verify that when soft sharding is enabled, Pods are preferentially scheduled to nodes within the assigned shard when sufficient resources are available. e2e need cover allocate, preempt, reclaim, backfill actions

• Soft Sharding Mode – Fallback to Out-of-Shard Nodes

Verify that when soft sharding is enabled and in-shard nodes do not have sufficient resources, Pods can be scheduled to nodes outside the shard as a fallback. e2e need cover allocate, preempt, reclaim, backfill actions

• Shard reassignment

Verify that after shard changes (e.g., add/remove node in nodeshard), Pods can be scheduled to different nodes according to the updated shard layout. e2e need cover allocate, preempt, reclaim, backfill actions

[aaradhychinche-alt]

/assign