kubeflow/sdk

[SDK] Get the correct TrainJob components using `get_job()` API

Open

#25 opened on Dec 11, 2024

View on GitHub
 (26 comments) (1 reaction) (3 assignees)Python (196 forks)auto 404
good first issuehelp wantedkind/feature

Repository metrics

Stars
 (124 stars)
PR merge metrics
 (PR metrics pending)

Description

What you would like to be added?

As we discussed, currently get_job() API can return multiple Pods for every TrainJob component, like initializer or trainer-node-0: https://github.com/kubeflow/training-operator/pull/2324#discussion_r1860773473. That can happen when Pods are re-created based on Batch/Job restart policies. Therefore, users can see unexpected logs while using the Kubeflow Training SDK.

We should improve this API to show the correct TrainJob components to users. For example, when we list all of the Pods, we can select the most recently created Pod with the same role (e.g. dataset-initializer).

Love this feature?

Give it a 👍 We prioritize the features with most 👍

Contributor guide

Research direction
Tech stack
Domain
Issue type
DifficultyEstimated implementation difficulty for a new contributor, from 1 for very small changes to 5 for expert-level work.
Estimated timeA rough time range for an experienced contributor to investigate, implement, test, and prepare a pull request.
Activity statusHow available the issue appears right now: fresh, active, stale, blocked, or waiting on maintainer input.
ClarityHow clearly the issue explains the expected change, acceptance criteria, and next step.
Prerequisites
Newbie friendlinessA 1-100 score estimating how approachable this issue is for first-time contributors.