docs(kep): draft fail-fast restart budget and init-phase DNS for LWS

[panpan0000]

Summary

• Add KEP draft for LWS fail-fast restart budget with terminal condition.
• Add optional support for init-phase peer DNS.
• Prioritize fail-fast as primary motivation; DNS support is secondary.

Key Points

• Add (opt-in, nil default).
• Add new LWS condition type .
• Keep current behavior unchanged when unset.
• Reuse existing LWS env vars; no new env introduced.

Notes

• This commit is docs-only and scoped to .

relevant to kubeflow/trainer#3417

Co-Author: GPT-5.4

[netlify]

✅ Deploy Preview for kubernetes-sigs-lws canceled.

Name	Link
🔨 Latest commit	758d435
🔍 Latest deploy log	https://app.netlify.com/projects/kubernetes-sigs-lws/deploys/6a1e80401f8b530008f3bb3c

[k8s-ci-robot]

[APPROVALNOTIFIER] This PR is NOT APPROVED

This pull-request has been approved by: panpan0000
Once this PR has been reviewed and has the lgtm label, please assign kerthcet for approval. For more information see the Code Review Process.

The full list of commands accepted by this bot can be found here.

Details

Needs approval from an approver in each of these files:

• OWNERS

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

[linux-foundation-easycla]

The committers listed above are authorized under a signed CLA.

• ✅ login: panpan0000 / name: Peter Pan (0590df4, 0d142bd, 758d435, f88b79a)

[panpan0000]

the 2nd commit( updating .github/workflows/ ) is due to the CI failure ... not sure if it works... just from AI....

[Edwinhr716]

@panpan0000 can you create an issue which describes what this KEP is trying to address?

[Edwinhr716]

Also to fix the failing tests, we need to wait for #814 to be completed

[panpan0000]

1. linked issue here How to Support distributed preflight checks in init phase #820
2. remove the CI run workaround and rebase origin/main

[panpan0000]

rebase to trigger CI again, after #814 merged, as @Edwinhr716 metioned.

[panpan0000]

itseems CI passed , please take a look when you have time, thank you! @Edwinhr716

[panpan0000]

ping

[panpan0000]

I saw another problem: @Edwinhr716

when vLLM pod crash and restarted by probe, nothing changed, except the AGE renews, silence failure....

NAME                                                            READY   STATUS                     RESTARTS       AGE
deepseek-v4-pro-llm-d-modelservice-decode-0                     0/1     Running                    0              5m20s
deepseek-v4-pro-llm-d-modelservice-decode-0-1                   1/1     Running                    0              5m20s

 kubectl -n public get lws -o wide
deepseek-v4-pro-llm-d-modelservice-decode   10d