test(trainer): expand unit coverage for trainer MCP tools

[abhijeet-dhumal]

Part of #9. Builds on the test scaffold in #6.
Cover trainer tool success and submission paths using mock_trainer_client and mock_k8s_apis

Scope

• training: confirmed=True paths for fine_tune, run_custom_training, run_container_training
• discovery: list/get jobs success paths, runtime listing edge cases
• monitoring: log/event parsing, failure-pattern hints
• lifecycle: suspend/resume/delete with mocked SDK
• platform: runtime CRUD preview and confirmed paths
• planning: pre_flight and estimate_resources with mocked cluster data
• common/utils: is_mcp_managed with mocked CustomObjects API

Out of scope

• Real cluster calls
• Integration/e2e
• Benchmarks (Benchmark suite for trainer MCP tools #10 / feat: add latency benchmarking and HTML reporting for trainer MCP tools #26)

Done when

• TODO(test) markers in tests/unit/trainer/ and related common/ cleared or tracked
• Pattern C tests (mock_trainer_client + verify_tool_*) for all write tools
• make test-python green

[abhijeet-dhumal]

/area testing
/area trainer
/good-first-issue

[google-oss-prow]

@abhijeet-dhumal:
This request has been marked as suitable for new contributors.

Please ensure the request meets the requirements listed here.

If this request no longer meets these requirements, the label can be removed
by commenting with the /remove-good-first-issue command.

Details

In response to this:

/area testing
/area trainer
/good-first-issue

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository.

[aastha-m22]

Hi @abhijeet-dhumal — following @humammoin09-blip's approach of splitting this up given the scope, I'd like to take the discovery section (list/get jobs success paths, runtime listing edge cases) using mock_trainer_client per the Pattern C convention. I recently wrote unit tests for dynamic_tools.py in this repo's sibling area (kubeflow/mcp-server #67), so I'm already familiar with the test conventions here. Let me know if that section's open!

[Solaris-star]

Hi! I would like to take the lifecycle section from this issue:

• suspend / resume / delete paths with mocked SDK
• following Pattern C (mock_trainer_client + verify_tool_*)

I will open a focused PR for just this slice so review stays small. Please let me know if that is already claimed!

[Solaris-star]

Also claiming the platform section from this issue as a separate focused PR (runtime CRUD preview/confirmed paths with mocks).

[Nagasirikancha]

Hi @abhijeet-dhumal! I'd like to take the training section from this issue, covering the confirmed=True paths for fine_tune, run_custom_training, and run_container_training, following the existing Pattern C convention (mock_trainer_client + verify_tool_*) used in #82, #95, and #97. I'll keep the PR focused on just this slice. Please let me know if it's still available. Thanks!

[ptalgulk01]

Hi @abhijeet-dhumal,I would like to pick up the monitoring section (log/event parsing, failure-pattern hints) and the common/utils section (is_mcp_managed with mocked CustomObjects API). Can please tell if they are available to work on ? Thankyou!