kubeflow/trainer

Enhanced MPI Support in TrainJob

Closed

#2,751 opened on Jul 24, 2025

View on GitHubContributor guide
 (26 comments) (3 reactions) (0 assignees)Go (964 forks)github user discovery
area/controllerarea/mpiarea/runtimesgood first issuehelp wantedkind/feature

Repository metrics

Stars
 (2,112 stars)
PR merge metrics
 (PR metrics pending)

Description

What you would like to be added?

We plan to enhance MPI support in TrainJob in these areas:

  • SSH-based node communication
  • Add resource per node support
  • Address the failing launcher issue in the initialization phase (launcher fails until the MPI worker is ready)

Why is this needed?

As we discussed in the WG call yesterday, this will be included in Kubeflow Trainer ROADMAP 2025.

/cc @kubeflow/wg-training-leads @astefanutti @franciscojavierarceo @kramaranya @rudeigerc @mahdikhashan @Doris-xm @Harshal292004 @saileshd1402 @kannon92 @deepanker13

Love this feature?

Give it a 👍 We prioritize the features with most 👍

Contributor guide

Research direction
Investigate the failing launcher issue in the initialization phase and understand the MPI worker startup sequence.
Tech stack
gokubernetes
Domain
machine learninginfrastructurebackend
Issue type
Feature
DifficultyEstimated implementation difficulty for a new contributor, from 1 for very small changes to 5 for expert-level work.
3
Estimated timeA rough time range for an experienced contributor to investigate, implement, test, and prepare a pull request.
Half day
Activity statusHow available the issue appears right now: fresh, active, stale, blocked, or waiting on maintainer input.
Active
ClarityHow clearly the issue explains the expected change, acceptance criteria, and next step.
Mostly clear
Prerequisites
GoKubernetes
Newbie friendlinessA 1-100 score estimating how approachable this issue is for first-time contributors.
50