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.

Sign up for GitHub

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

Training: Add documentation for the MultiKueue and spec.managedBy API #3956

Open
wants to merge 7 commits into
base: master
Choose a base branch
from
Open

Training: Add documentation for the MultiKueue and spec.managedBy API #3956

Changes from 6 commits
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
4b02eec
resolves kubeflow/training/#2279
Garvit-77 Jan 13, 2025
fe517b1
Update content/en/docs/components/training/user-guides/managedby.md
Garvit-77 Jan 23, 2025
af5ac9e
Update content/en/docs/components/training/user-guides/managedby.md
Garvit-77 Jan 23, 2025
2c2cd7d
Update content/en/docs/components/training/user-guides/managedby.md
Garvit-77 Jan 23, 2025
cbac0ed
Update content/en/docs/components/training/user-guides/managedby.md
Garvit-77 Jan 23, 2025
713beb7
Update content/en/docs/components/training/user-guides/managedby.md
Garvit-77 Jan 23, 2025
0df155e
Update content/en/docs/components/training/user-guides/managedby.md
Garvit-77 Jan 24, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
65 changes: 65 additions & 0 deletions content/en/docs/components/training/user-guides/managedby.md
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,65 @@
+++
title = "spec.managedBy"
Desciption = "Using managedBy field for MultiKueue"
weight = 60
+++

## Overview

The `spec.managedBy` field is a new feature introduced in the Kubeflow Training Operator to support a more robust multi-cluster job dispatching by [MultiKueue](https://kueue.sigs.k8s.io/docs/concepts/multikueue/).
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

small adjustment spec.runPolicy.managedBy

Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

+1, in Kubeflow 1.9 this is under spec.runPolicy: https://github.com/kubeflow/trainer/blob/078ec30ff26649a07d2e28893c332d4cef70e233/pkg/apis/kubeflow.org/v1/common_types.go#L231C5-L239. Also adjust the code snippets to use

spec:
  runPolicy:
     managedBy: "kueue.x-k8s.io/multikueue"


## Prerequisites

1. Ensure that you have the latest version of the Kubeflow Training Operator installed.
2. Make sure Kueue is compiled against the new operator to leverage the `spec.managedBy` field.

## Usage

To use the `spec.managedBy` field in your training jobs, include it in the job specification as shown below:

```yaml
apiVersion: "kubeflow.org/v1"
kind: "TFJob"
metadata:
name: "example-tfjob"
spec:
managedBy: "kueue.x-k8s.io/multikueue"
tfReplicaSpecs:
...
```

Example
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Actually, underneath Kueue will add a webhook which will default the field, so the example will not need to include it. I think we can just rely on the example eventually added to the MultiKueue documentation once kubernetes-sigs/kueue#2552 is done.

Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

WDYT @tenzen-y @andreyvelich ?

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes, I think we can reference to the example in Kueue docs, so we can be consistent.

Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

ok, I suggest to drop it in this PR so that we only have one place about the details. Note that the Kueue documentation is not added yet. Once done we will x-ref the pages.


Here is a complete example of a TensorFlow job using the spec.managedBy field:

```YAML
apiVersion: "kubeflow.org/v1"
kind: "TFJob"
metadata:
name: "example-tfjob"
spec:
managedBy: "kueue"
tfReplicaSpecs:
Chief:
replicas: 1
template:
spec:
containers:
- name: tensorflow
image: tensorflow/tensorflow:latest
args: ["python", "model.py"]
Worker:
replicas: 2
template:
spec:
containers:
- name: tensorflow
image: tensorflow/tensorflow:latest
args: ["python", "model.py"]
```

## What's next?

For more details on setting up and using MultiKueue with the Kubeflow Training Operator, refer to the following documentation pages:

- [Kueue/Kubeflow](https://kueue.sigs.k8s.io/docs/tasks/run/multikueue/kubeflow/)