如何在Slurm中定义DCU设备?
我的定义如下:
slurm.conf 定义:
GresTypes=gpu,dcu,mlu
slurm_node.conf 定义:
NodeName=compute05 NodeHostname=compute05 NodeAddr=compute05 CPUs=128 Boards=1 SocketsPerBoard=8 CoresPerSocket=8 ThreadsPerCore=2 RealMemory=128649 MemSpecLimit=4096 Feature="DCU Z100L" Gres=dcu:z100:8
gres.conf 定义:
NodeName=compute05 Name=dcu Type=z100 File=/dev/dri/card0
NodeName=compute05 Name=dcu Type=z100 File=/dev/dri/card1
NodeName=compute05 Name=dcu Type=z100 File=/dev/dri/card2
NodeName=compute05 Name=dcu Type=z100 File=/dev/dri/card3
NodeName=compute05 Name=dcu Type=z100 File=/dev/dri/card4
NodeName=compute05 Name=dcu Type=z100 File=/dev/dri/card5
NodeName=compute05 Name=dcu Type=z100 File=/dev/dri/card6
NodeName=compute05 Name=dcu Type=z100 File=/dev/dri/card7
结果DCU节点总是drain,显示
gres/dcu count reported lower than configured (0 < 8)
那么我该如何定义dcu设备呢?

暂无数据