WingKai Ho comments

Results 7 comments of


                                            WingKai Ho

Use nvidia-smi to ban error gpu make UnexpectedAdmissionError with failed to find gpu id, which is unexpected

https://github.com/volcano-sh/devices/pull/29

Use nvidia-smi to ban error gpu make UnexpectedAdmissionError with failed to find gpu id, which is unexpected

I think space of volcano-schduler gpu ids are 0 - GpuDeviceCount - 1; but if you use nvidai-smi to ban error gpu, the gpu ids space of is not start...

Use nvidia-smi to ban error gpu make UnexpectedAdmissionError with failed to find gpu id, which is unexpected

@HariVamsiK it fix in https://github.com/volcano-sh/devices/pull/29

Use nvidia-smi to ban error gpu make UnexpectedAdmissionError with failed to find gpu id, which is unexpected

/assign @Thor-wl

fix gpu shareing predictor allocate more than one gpu for pod

/assign @thor-wl

fix predicateGPUbyMemory: sort of devIDs reduce fragment gpu

Sorry，last time forget sort devIDs before return. https://github.com/volcano-sh/volcano/pull/2465

Volcano gpu-share plugin patch all suitalbe for pod

https://github.com/volcano-sh/volcano/blob/master/pkg/scheduler/plugins/predicates/predicates.go#L197 retrun all suitable gpu ids for pod. But pod need only one gpu id.