kubeflow / sdk / 29323902563

84%

push

github

fix(trainer): don't report Complete when all container statuses are Unknown (#562)

aggregate_status_from_containers() ran all() over statuses filtered to
exclude UNKNOWN. When the list is empty or every status is UNKNOWN the
filtered sequence is empty, so all([]) returns True and the job is
reported Complete. This is reachable via aggregate_container_statuses()
when a container cannot be inspected (get_container_status returns
UNKNOWN), so an unreadable job wrongly reports success to get_job(),
list_jobs(), wait_for_job_status() and get_job_logs().

Require at least one known status and all known statuses to be Complete
before returning Complete; empty or all-UNKNOWN now aggregate to Unknown.
Add unit test coverage for aggregate_status_from_containers.

Fixes #561

Signed-off-by: Sahil Kumar Singh <sahilkumargreat12@gmail.com>

10 of 10 new or added lines in 2 files covered. (100.0%)

5276 of 6293 relevant lines covered (83.84%)

1.68 hits per line