NVIDIA / holodeck / 29180595118
52%
main: 51%

LAST BUILD BRANCH: pull-request/856
DEFAULT BRANCH: main
Ran
Jobs 1
Files 84
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: 51.879%. Remained the same
29180595118

push

github

ArangoGutierrez 
fix(e2e): harden real-ssh harness against container-startup flake

PR #856's first CI run failed when a container crashed at boot: --rm
reaped the crashed container out from under installBash's blind 30x1s
retry loop, which spent 30 seconds getting daemon 404s against a
container ID that no longer existed, and no docker logs survived to
diagnose the crash.

- startSSHContainer no longer runs with --rm; DeferCleanup now uses
  'docker rm -f' (kills and removes, so no orphan is left regardless
  of which attempt succeeds).
- installBash checks 'docker inspect -f {{.State.Running}}' before
  each apk-add attempt. A non-running container fails fast with the
  inspect state and the tail of 'docker logs' instead of exhausting
  the 30x1s budget against a dead container; the same log tail is
  attached to the never-succeeded failure path too.
- Container start + installBash is now one retryable attempt
  (attemptStartSSHContainer). startSSHContainer removes a failed
  first container via 'docker rm -f' and retries once with a fresh
  container, logging a By() line so CI output shows the flake fired
  even when self-healed; a second consecutive failure surfaces the
  full inspect-state + docker-logs evidence via Fail().

Image ref, env vars, port publishing, waitForSSH, and the spec body
are unchanged.

Signed-off-by: Carlos Eduardo Arango Gutierrez <eduardoa@nvidia.com>

6833 of 13171 relevant lines covered (51.88%)

0.58 hits per line

Jobs
ID Job ID Ran Files Coverage
1 29180595118.1 84
51.88
 GitHub Action Run
Source Files on build 29180595118