Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Add environment name in ControllerConfig #450

Closed
wbuchwalter opened this issue Mar 8, 2018 · 6 comments
Closed

Add environment name in ControllerConfig #450

wbuchwalter opened this issue Mar 8, 2018 · 6 comments

Comments

@wbuchwalter
Copy link
Contributor

It would be useful to add the environment name (gke, azure etc.) in the ControllerConfig.

For example, in the dashboard we want to be able to specify the resources.limits for master/worker.
However for GPUs, GKE is using the device plugin, so nvidia.com/gpu while Azure is using alpha.kubernetes.io/nvidia-gpu for example.
The dashboard could first read the environment name in the ControllerConfig first and choose the appropriate resource based on this.

@jlewi thoughts?

@jimexist
Copy link
Member

jimexist commented Mar 9, 2018

@wbuchwalter did you mean #451 ?

@wbuchwalter
Copy link
Contributor Author

wbuchwalter commented Mar 9, 2018

@jimexist Yes :) Although, not sure if the GetGPUResourceName should be where you put it.
Unless it has other possible use someplace else, I was thinking doing that it the front-end, because there might be a bunch of different customization we might want to do in the dashboard based on the cloud.

@jlewi
Copy link
Contributor

jlewi commented Mar 17, 2018

Would it be better to use the ksonnet template as opposed to the controller config to handle GPU config?

In the way we have Cloud specific parts for TFServing
https://github.com/kubeflow/kubeflow/blob/master/kubeflow/tf-serving/tf-serving.libsonnet

We could have cloud specific bits for TFJob and we could use this to specify resources like the GPU resource name conditioned on the cloud.

Using the operator to specify GPU config predates our use of ksonnet. Now that we are using ksonnet it might make more sense to use ksonnet to define suitable templates for different clouds?

@jlewi
Copy link
Contributor

jlewi commented May 8, 2018

@wbuchwalter Any update on this? Should the different resource names just be handled in the UI?

E.g. in the UI we could let users specify alpha.kubernetes.io/nvidia-gpu or nvidia.com/gpu? They could either specify that directly or they could specify cloud and then the UI would infer the values based on this.

What priority should we assign this issue? If it must get done and will get done for our 0.2 release (end of June) then it should be P1 otherwise P2.

@wbuchwalter
Copy link
Contributor Author

Haven't worked on this yet, but let's mark it as P1.

@stale
Copy link

stale bot commented Apr 20, 2020

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

@stale stale bot closed this as completed Apr 27, 2020
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

No branches or pull requests

4 participants