31529637572
68%

Ran 11 Aug 2026 07:52PM UTC

Jobs 1

Files 42

Run time 1min

Badge

Embed ▾

Committed 11 Aug 2026 07:47PM UTC coverage: 68.718% (+0.2%) from 68.536%

Build # 31529637572

Build Type

push

github

Committed by

web-flow

Commit Message

fix(operator): keep the TrainJob Failed condition terminal (#3891)

The reconciler cleared the Failed condition at the top of every loop, so a
TrainJob that had already failed looked unfinished for the rest of the pass.
With the JobSet gone, IsTrainJobFinished returned false and reconcileObjects
recreated it, resetting status.restarts and letting maxRestarts run forever.

The condition was added to let an external change to the TrainJob spec move
it back out of Failed, but there is no longer a path to do that. The only
Failed the reconciler itself sets outside the deadline check is
TrainingRuntimeNotSupported, and recovering from it would mean pointing
runtimeRef at a registered GroupKind. runtimeRef is immutable, and the
runtime registry is injected when the reconciler is built.

Remove the clearing so Failed stays terminal.

Signed-off-by: XploY04 <2004agarwalyash@gmail.com>

Coverage Stats

2579 of 3753 relevant lines covered (68.72%)

0.8 hits per line

Jobs

ID	Job ID	Ran	Files	Coverage
1	31529637572.1	11 Aug 2026 07:52PM UTC	42	68.72	GitHub Action Run

kubeflow / trainer / 31529637572
68%

README BADGES
x

Markdown

Textile

RDoc

HTML

Rst

Jobs

Source Files on build 31529637572

kubeflow / trainer / 31529637572 68%

README BADGES x

Markdown

Textile

RDoc

HTML

Rst

Jobs

Source Files on build 31529637572

kubeflow / trainer / 31529637572
68%

README BADGES
x