limengxuan

vgpu support for volcano

5

resubmit this pr to merge into master branch of volcano vgpu device plugin repo: https://github.com/archlitchi/devices

Who's using HAMi / 您在使用 HAMi 吗 ?

53

Sincerely thank you for using and continuing to pay attention to HAMi. In order to better build the community and attract more people to use and contribute to HAMi to...

move device-sharing logics from predicates to standalone devices plugin

2

In order support other Heterogeneous AI Computing devices, we'd better separate device-sharing logics to a standalone plugin called 'devices'

Add nodeOrder for devices API and implement idle-first fit and best-fit for vgpu

10

1. Add nodeOrder plugin for future scheduling policies(best fit, first fit ,etc..) 2. Add vgpu monitor and schedule policy related document in how_to_use_vgpu.md

Unexpected Admission error while using vGPU with vc-jobs replicas>=2

1

Step to reproduce: 1. Enable vGPU 2. submit the following yaml ``` apiVersion: batch.volcano.sh/v1alpha1 kind: Job metadata: name: lm-mpi-job spec: minAvailable: 3 schedulerName: volcano plugins: ssh: [] svc: [] tasks:...

Redraw the architect figure

2

Current figure: https://github.com/Project-HAMi/HAMi/blob/master/imgs/arch.png is not very beautiful Could anyone redraw this figure to make it more professional?

fix allocating more pods to a GPU when using volcano-vgpu feature

3

fix an issue: allocate more pods to a GPU resulting a GPU exceeds its core limit.  related issue: https://github.com/Project-HAMi/volcano-vgpu-device-plugin/issues/30