Configuring Scheduling Plugins

If resource overcommitment is not required, ensure that the Volcano scheduling plugins do not contain the overcommit plugin.

If - name: overcommit exists in the Volcano deployment file volcano-v*.yaml, delete it.

...
data:
  volcano-scheduler.conf: |-
    actions: "enqueue, allocate, backfill"
    tiers:
    - plugins:
      - name: volcano-npu-v3.0.0_linux-aarch64  # v3.0.0 indicates the MindX DL version. The number varies depending on the version.
    - plugins:
      - name: priority
      - name: gang
      - name: conformance
    - plugins:
      - name: drf
      - name: predicates
      - name: proportion
      - name: nodeorder
      - name: binpack
    configurations:
      - name: selector
        arguments: {"host-arch":"huawei-arm|huawei-x86",
        "accelerator":"huawei-Ascend910|nvidia-tesla-v100|nvidia-tesla-p40",
        "accelerator-type":"card|module|half","servertype":"soc"}
      - name: init-params
        arguments: {"grace-over-time":"900","presetVirtualDevice":"true"}
...

If the overcommit plugin exists, it accepts jobs by occupying 1.2 times of resources. In this way, the job throughput and bandwidth are improved. However, after jobs with overloaded resources enter the scheduling queue, resources cannot be allocated to schedulable jobs.