Enable link training support in sonic-mgmt

[sdszhang]

Description of PR

Summary:
Enable link-training support in sonic-mgmt

Currently, enabling them for all server facing ports in F2 T0 topo. Once we have port-channel support, we can enable them only for port-channel in F2 topo.

Type of change

• Bug fix
• Testbed and Framework(new/improvement)
• New Test case
  • Skipped for non-supported platforms
• Test case improvement

Back port request

• 202205
• 202305
• 202311
• 202405
• 202411
• 202505
• 202511

Approach

What is the motivation for this PR?

Enable link-training support in sonic-mgmt

How did you do it?

Enabling link training in golden_config_db for F2 topology.

How did you verify/test it?

local testbed.

Any platform specific information?

Supported testbed topology if it's a new test case?

Documentation

[mssonicbld]

/azp run

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[mssonicbld]

/azp run

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[wsycqyz]

Hi @sdszhang, does the 202511 branch need this change as well?

[mssonicbld]

/azp run

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[sdszhang]

Hi @sdszhang, does the 202511 branch need this change as well?

Yes. 202511 will need this too.

[mssonicbld]

/azp run

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[Copilot]

Pull request overview

This PR adds link-training configuration support to sonic-mgmt, primarily targeting the t0-f2 topology by enabling link training on server-facing ports via golden_config generation, and by allowing link-training settings to flow from testbed link CSVs into fanout SONiC deployment configs.

Changes:

• Add optional LinkTraining parsing in the Ansible graph CSV processing and propagate it into per-link facts.
• Extend SONiC fanout deployment templates (202405/202505) to emit PORT.link_training based on connection metadata.
• Add a new golden config generator path for t0-f2 to enable PORT.link_training on server-facing ports.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 4 comments.

File	Description
ansible/roles/fanout/templates/sonic_deploy_202505.j2	Emit link_training in generated fanout config; broaden FEC “rs” speed condition.
ansible/roles/fanout/templates/sonic_deploy_202405.j2	Emit link_training in generated fanout config; add 200G to FEC “rs” speed condition.
ansible/module_utils/graph_utils.py	Parse optional LinkTraining column from links CSV into graph facts.
ansible/library/generate_golden_config_db.py	Add t0-f2 golden config generation to enable link training on server-facing ports.

---

You can also share your feedback on Copilot code review. Take the survey.

[mssonicbld]

/azp run

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[mssonicbld]

/azp run

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[StormLiangMS]

Code Review

✅ LGTM — Good addition of link training support for F2 topology across golden_config generation, graph_utils CSV parsing, and fanout deploy templates (both 202405 and 202505).

Minor suggestion: In generate_t0_f2_golden_config_db():

ori_config = json.loads(self.get_config_from_minigraph())
golden_config = ori_config  # This is a reference, not a copy

Consider using copy.deepcopy(ori_config) for clarity. Currently golden_config and ori_config are the same object, so mutations apply to both. It works, but could confuse future maintainers who expect them to be independent.

[mssonicbld]

/azp run

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[StormLiangMS]

Approve with minor comments ✅

Overall this is a solid feature addition enabling link-training for F2 T0 topology. The 4-file change covers the full stack: CSV parsing → graph facts → golden config generation → fanout deployment.

Comments

1. generate_t0_f2_golden_config_db() — reference vs. copy (nit)

golden_config = ori_config          # reference, not copy
golden_config["PORT"] = ori_config.get("PORT", {})  # no-op, same dict

This works because ori_config is freshly parsed and never reused, but it reads confusingly. Consider golden_config = {"PORT": ori_config.get("PORT", {})} directly, or use copy.deepcopy for clarity.

2. Fanout template indentation (cosmetic)
In sonic_deploy_202405.j2, "link_training": "on" renders at 16 spaces indentation while other PORT keys ("fec", "alias", etc.) render at 12 spaces. Not a bug (JSON parsers ignore whitespace), but inconsistent. The pre-existing "autoneg" block at line 33 has the same issue (20 spaces).

3. FEC speed list change in sonic_deploy_202505.j2 (notable)
The FEC condition was silently broadened from == "100000" to in ["100000", "200000", "400000", "800000"]. This is a behavior change beyond link-training — now 200G/400G/800G ports on fanouts will also get "fec": "rs". Worth calling out in the PR description since it affects all fanout deployments, not just link-training ones.

4. 202405 vs 202505 template asymmetry
The 202505 template handles both link_training: "on" and "off", but the 202405 template only handles "on". If a CSV entry has LinkTraining=off, it won''t be set in the 202405 fanout config. This may be intentional (202405 default is off), but worth noting.

None of these are blockers — all are minor/cosmetic.

[mssonicbld]

@sdszhang PR conflicts with 202511 branch