zzz comments

Results 46 comments of

zzz

Support setting the timeout of guest OCI hooks

@liubin not yet, and I'll do it this week.

WIP: runtime: Add support for GPUDirect and GPUDirect RDMA PCIe topology Part 2

thanks @zvonkok, let me test it when it is ready.

WIP: runtime: Add support for GPUDirect and GPUDirect RDMA PCIe topology Part 2

@zvonkok that would be fantastic, and let me test it. BTY, is the `--label` parameter required? In my case, I need to use it in combination with k8s. ``` sudo...

runtime: fix PCIe topology for GPUDirect use-case

> @soulseen With the limitations of QEMU you are only allowed to hot-plug 4 GPUs with 4 devices in each IOMMU-Group meaning a total of 16 devices is theoretically possible....

Cannot find package "." while updating k8s.io/client-go

Is there any solution?

Support backend storage driver: Aliyun OSS.

@bitsf about Access id not inject, may I bother you for sharing your Ali OSS secret yaml?

AddAfter should be able to cancel previous enqueues of the same key

/remove-lifecycle stale

AddAfter should be able to cancel previous enqueues of the same key

Any progress about this issue?

support hotplug/hot-unplug VFIO Device in container running state

@fengwang666 yes, the GPU hotplug/unplug lifecycle is independent of the pod/containers. in the serverless use case, we have some applications that only need the GPU at certain times, and for...

support hotplug/hot-unplug VFIO Device in container running state

@bergwolf yes, as you said, in the view of decoupling, GPU resources will be independent of k8s, and they will be controlled by other components that will hotplug GPU to...