You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@yunikorn.apache.org by "TingYao Huang (Jira)" <ji...@apache.org> on 2026/09/28 17:48:00 UTC
[jira] [Resolved] (YUNIKORN-3480) Weekly e2e fails on missing kind node images and a merged 1.37/1.36 matrix entry
[ https://issues.apache.org/jira/browse/YUNIKORN-3480?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
TingYao Huang resolved YUNIKORN-3480.
-------------------------------------
Fix Version/s: 1.11.0
Target Version: 1.11.0
Resolution: Fixed
> Weekly e2e fails on missing kind node images and a merged 1.37/1.36 matrix entry
> --------------------------------------------------------------------------------
>
> Key: YUNIKORN-3480
> URL: https://issues.apache.org/jira/browse/YUNIKORN-3480
> Project: Apache YuniKorn
> Issue Type: Bug
> Components: test - e2e
> Reporter: Dian-Xuan Yang
> Assignee: Dian-Xuan Yang
> Priority: Major
> Labels: pull-request-available
> Fix For: 1.11.0
>
>
> Since YUNIKORN-3455 (k8shim PR #1096, merged 2026-09-18), the weekly e2e workflow has failed on every scheduled run: 4 of its 13 jobs failed in both the 2026-09-20 and 2026-09-27 runs. The 2026-09-13 run, before that change, did not have these failures. There are two independent defects in {{.github/workflows/weekly-e2e.yml}}.
> h3. 1. A missing comma merges two matrix entries
> {code:yaml}
> k8s:
> [
> v1.37.0
> v1.36.4,
> v1.35.8,
> {code}
> Without the comma after {{v1.37.0}}, YAML parses these two entries as a single scalar, {{v1.37.0 v1.36.4}}. As a result, the job attempts to run {{docker pull 'kindest/node:v1.37.0 v1.36.4'}} and fails:
> {noformat}
> ERROR: failed to create cluster: failed to pull image "kindest/node:v1.37.0 v1.36.4":
> command "docker pull 'kindest/node:v1.37.0 v1.36.4'" failed with error: exit status 1
> {noformat}
> The impact is broader than a single failed job: the matrix contains 14 versions but expands to only 13 entries. As a result, *neither 1.37.0 nor 1.36.4 is actually exercised by the weekly run*, which defeats the purpose of adding these versions in YUNIKORN-3455.
> h3. 2. Three kind node images do not exist
> The version bump followed the Kubernetes patch releases rather than the node image versions actually published by kind. As a result, three entries return 404 from Docker Hub:
> ||matrix entry||exists||latest image kind publishes for that minor||
> |v1.33.13|no|v1.33.12|
> |v1.32.13|no|v1.32.11 (the value this change replaced)|
> |v1.30.14|no|v1.30.13|
> {noformat}
> ERROR: failed to create cluster: failed to pull image "kindest/node:v1.33.13":
> command "docker pull kindest/node:v1.33.13" failed with error: exit status 1
> {noformat}
> The remaining ten entries (v1.35.8, v1.34.11, v1.31.14, v1.29.14, v1.28.15, v1.27.16, v1.26.15, v1.25.16, v1.24.17) were checked and do exist.
> {{.github/workflows/pre-commit.yml}} is not affected: its matrix includes the missing comma and lists only published images. {{branch-1.10}} still carries the pre-bump matrix, so the issue is limited to master.
> h3. Fix
> Add the missing comma and pin the three invalid entries to the kind node images that are actually published.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@yunikorn.apache.org
For additional commands, e-mail: dev-help@yunikorn.apache.org