nvidia/k8s-device-plugin:1.0.0-beta3 image is missing from docker hub

[jucrouzet]

$ docker pull nvidia/k8s-device-plugin:1.0.0-beta3
Error response from daemon: manifest for nvidia/k8s-device-plugin:1.0.0-beta3 not found: manifest unknown: manifest unknown

[jucrouzet]

When I build it, i get :

panic: runtime error: invalid memory address or nil pointer dereference
[signal SIGSEGV: segmentation violation code=0x1 addr=0x0 pc=0x8414d7]

goroutine 1 [running]:
main.getDevices(0xc4200c3e38, 0x1, 0x1)
/go/src/nvidia-device-plugin/nvidia.go:49 +0x177
main.main()
/go/src/nvidia-device-plugin/main.go:42 +0x111

[RenaudWasTaken]

I just restarted the pipeline, the image should be there in a few minutes or so.
@klueska looks like nvml has a Segfaults on here:

k8s-device-plugin/nvidia.go

Line 49 in 0b12465

ID: int64(*(d.CPUAffinity)),

[rockrush]

@jucrouzet @RenaudWasTaken Pull request #141 fixes the problem(hopefully), please try and verify.

[klueska]

@RenaudWasTaken I left a comment on #141

If d.CPUAffinity is Nil then we should just not set the Topology field of the device at all.

When is it the case that d.CPUAffinity is not set, such that this occurs?

[zplizzi]

I'm getting the same error.

I installed the plugin as recommended on the main Readme.md with
kubectl create -f https://raw.githubusercontent.com/NVIDIA/k8s-device-plugin/1.0.0-beta3/nvidia-device-plugin.yml

Here's the logs for the pod running on an AWS p3.2xlarge instance:

% kubectl logs -n kube-system nvidia-device-plugin-daemonset-4h5pl
2019/10/11 19:06:30 Loading NVML
2019/10/11 19:06:30 Fetching devices.
2019/10/11 19:06:30 Shutdown of NVML returned: <nil>
panic: runtime error: invalid memory address or nil pointer dereference
[signal SIGSEGV: segmentation violation code=0x1 addr=0x0 pc=0x8414d7]

goroutine 1 [running]:
main.getDevices(0xc42006fe38, 0x1, 0x1)
        /go/src/nvidia-device-plugin/nvidia.go:49 +0x177
main.main()
        /go/src/nvidia-device-plugin/main.go:42 +0x111

Is there a recommended version to install instead that should work?

[zplizzi]

Dropping back to the previous version seems to work. Uninstalled broken version with

kubectl delete daemonset nvidia-device-plugin-daemonset -n kube-system

and installed previous version (beta instead of beta3) with

kubectl create -f https://raw.githubusercontent.com/NVIDIA/k8s-device-plugin/1.0.0-beta/nvidia-device-plugin.yml

[klueska]

This issue should be fixed by #141. No release has been made including this fix yet though.

[klueska]

Fix released: https://github.com/NVIDIA/k8s-device-plugin/releases/tag/1.0.0-beta4