GPU support

[enxebre]

/kind feature

Describe the solution you'd like
As a user I'd like to declarative define which GPU I want for my instances

Anything else you would like to add:
https://cloud.google.com/compute/docs/gpus/add-gpus#create-gpu-instance-api

We'd need to expose this in the machineInfra API and do the plumbing with GCP API through:

// GuestAccelerators: A list of the type and count of accelerator cards
// attached to the instance.
GuestAccelerators []*AcceleratorConfig `json:"guestAccelerators,omitempty"`

https://raw.githubusercontent.com/googleapis/google-api-go-client/master/compute/v1/compute-gen.go

[fejta-bot]

Issues go stale after 90d of inactivity.
Mark the issue as fresh with /remove-lifecycle stale.
Stale issues rot after an additional 30d of inactivity and eventually close.

If this issue is safe to close now please do so with /close.

Send feedback to sig-testing, kubernetes/test-infra and/or fejta.
/lifecycle stale

[detiber]

/lifecycle frozen

[dilyevsky]

I would like to claim that feature - need gpu support for GCP provider in our clusters.

[cpanato]

go for it @dilyevsky 🎉

[SamuelStuchly]

@dilyevsky Any progress on this feature ? If not I could take it, since we have already implemented this for Openshift at openshift/cluster-api-provider-gcp#172

[sayantani11]

@SamuelStuchly You can do the necessary changes.... And that will be the best, and then I can proceed with the documentation for gpu-support😄

[kernelpanic77]

Hello @cpanato @dims @richardcase!
I am Ishan from India. I am currently working in the computer vision domain, particularly in Medical-AI. Along with Distributed Systems, I also have knowledge of machine learning. I found the Kubernetes-sigs project extremely interesting and wanted to contribute to adding GPU support as Part of LFX for Summer 2022. I am currently going through the documentation to get a better understanding of the codebase. The LFX website shows that mentee applications for this mentorship are closed. Please let me know if I can still submit my application for the project.

Looking forward to contributing to this amazing project.

[reyvonger]

any updates?

[nicolas2bonfils]

Hello,

3 related PR were closed (#529, #655 and #654) due to inactivity.
What can we do to move forward on GPU support ? Are people interested by working on it (I wiling to participate) ?
Maybe openshift version of this provider can help ?

Regards

[nicolas2bonfils]

@SamuelStuchly do you think that your code in the Openshift version can also be used for this provider ?

[gecube]

A very important and necessary thing

[nicolas2bonfils]

I look at the code and it's very different from the version in the openshift PR. A big work is ahead of us ! 💪

[reyvonger]

bump