[Bug] Error when create bulk users

[edenbuaa]

OpenPAI Environment:

• OpenPAI version: v1.7.0 or v1.3.0
• Cloud provider or hardware configuration:
• OS (e.g. from /etc/os-release):
• Kernel (e.g. uname -a):
• Hardware (e.g. core number, memory size, storage size, GPU type etc.):
• Others:

Anything else we need to know:

openpai webportal report error when click 'Create Bulk Users'

[yiyione]

Hi @edenbuaa, I think you click the Create Bulk Users button in host/user-view.html page. And after click the button, will be navigated to host/batch-register.html page. Can you show which page you got the error above?

[edenbuaa]

@yiyione Yes, The page nvigated to batch-register.html page after click the Create Bulk users button. But when I click the Add New , that error arise.

[yiyione]

OK, I got the same error, it will be fix in next release. And workaround for the batch-register feature, please use the Download template, add the users into the csv file, and click Browser... button to import it.

[edenbuaa]

@Binyang2014 Is there a api to get the gpu count by sku in vc?

[Binyang2014]

I believe there is only one SKU in a vc. So you can get the gpu count in a VC via https://cluster-name/api/v2/virtual-clusters

[edenbuaa]

@Binyang2014 Maybe not, There is more one SKUs in a vc.

[edenbuaa]

@Binyang2014 The --cpu-manager-policy feature doesn't work for job pod. Is there a way to fix it?

Reproduce steps:

1. enable the --cpu-manager-policy feature in kubelet
2. submit a job
3. check the cpusets allocated for pod(/var/lib/kubelet/cpu_manager_state), but no entries.

[Binyang2014]

Can you paste the full kubelet configuration? You need to set --cpu-manager-policy="static". The default value is none. BTW PAI doesn't hack for kubelet, so you will get more help from k8s community.