kubeflow / mcp-server / 39473422536
87%

DEFAULT BRANCH: main
Ran
Jobs 3
Files 38
Run time 1min
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst
coverage: 85.996% (+0.1%) from 85.887%
39473422536

push

github

web-flow 
feat(trainer): report per-step status in get_training_job (#425)

* feat(trainer): report per-step status in get_training_job

get_training_job only returned name, status and runtime, although the SDK
TrainJob already carries each step (initializers and node-N pods) with its pod
phase, device and device count, plus num_nodes. An agent had to pull logs or
events to find out which pod was pending or had failed.

Add steps (sorted by name, at most MAX_JOB_STEPS), step_summary (counts by
status over all steps, plus a truncated flag) and num_nodes to the response.
Existing fields are unchanged. next_steps now points at the failed step's logs,
or at events when pods are pending. A step with no phase yet is Unknown.

Unit tests cover mixed step states, a failed and a pending step, a job without
steps, and the bound. The Kind E2E monitoring test checks the response shape
on a live job.

Part of #310

Signed-off-by: pratik naik <naikp0241@gmail.com>

* fix(trainer): keep failed steps in get_training_job when capping

Review feedback on #425: steps were sorted by name before the 32-step cap, so
in a large job a failed worker could be cut from steps, and next_steps would
not point at its logs while step_summary still reported the failure. Sort
failed steps first, then by name, and test a failed step past the cap.

Signed-off-by: pratik naik <naikp0241@gmail.com>

---------

Signed-off-by: pratik naik <naikp0241@gmail.com>

2788 of 3242 relevant lines covered (86.0%)

2.58 hits per line

Coverage Regressions

Lines Coverage ∆ File
36
81.14
2.09% trainer/api/discovery.py
Jobs
ID Job ID Ran Files Coverage
1 python-3.12 - 39473422536.1 38
86.0
GitHub Action Run
2 python-3.11 - 39473422536.2 38
86.0
GitHub Action Run
3 python-3.10 - 39473422536.3 38
86.0
GitHub Action Run
Source Files on build 39473422536