Skip to content

Fix lang-SDK k8s system test dropping the api-server auth manager#69221

Merged
bugraoz93 merged 1 commit into
apache:mainfrom
potiuk:fix-lang-sdk-k8s-auth-manager
Jul 1, 2026
Merged

Fix lang-SDK k8s system test dropping the api-server auth manager#69221
bugraoz93 merged 1 commit into
apache:mainfrom
potiuk:fix-lang-sdk-k8s-auth-manager

Conversation

@potiuk

@potiuk potiuk commented Jul 1, 2026

Copy link
Copy Markdown
Member

The KubernetesExecutor-3.13 k8s system test job started failing on main
after #68709 with 9 FileNotFoundError: /tmp/simple_auth_manager_passwords.json.generated
errors (every API-login test), e.g.
https://github.com/apache/airflow/actions/runs/28526595970/job/84584069072

Root cause: the lang-SDK provisioning (breeze k8s setup-lang-sdk-test)
helm-upgrades the already-deployed release with only lang_sdk/config/values.yaml.
Without --reuse-values, Helm re-renders from chart defaults and drops the base
deploy's --set overrides — notably config.core.auth_manager=SimpleAuthManager
(set on Python 3.13, where FAB is unavailable). The api-server reverts to the
chart-default FabAuthManager and never writes the generated password file the
tests fetch.

Fix: add --reuse-values so the lang-SDK values layer on top of the existing
release instead of replacing it — matching the intent (the same PR already made
dagBundleConfigList additive for exactly this reason).

The change is only exercised by the K8s system test job itself (no unit-test
harness exists for these breeze helm-command builders), so verification is that
job going green.

related: #68709


Was generative AI tooling used to co-author this PR?
  • Yes — Claude Code (Opus 4.8 1M)

Generated-by: Claude Code (Opus 4.8 1M) following the guidelines

… manager

The lang-SDK coordinator system test provisioning helm-upgrades the
already-deployed Airflow release with only the lang-SDK values file. Because
the upgrade did not reuse the release's existing values, Helm re-rendered from
chart defaults and discarded the base deploy's --set overrides -- including
config.core.auth_manager, which the base deploy sets to SimpleAuthManager on
Python 3.13 (FAB does not support 3.13). The api-server then came back up on
the chart-default FabAuthManager and never wrote
simple_auth_manager_passwords.json.generated, so every API-login test in the
KubernetesExecutor-3.13 variant errored with FileNotFoundError while fetching
the generated admin password.

Reuse the existing release's values on the lang-SDK upgrade so the overrides
layer on top and the auth manager (and the rest of the base configuration)
stays intact.
@potiuk potiuk added the all versions If set, the CI build will be forced to use all versions of Python/K8S/DBs label Jul 1, 2026
@potiuk potiuk closed this Jul 1, 2026
@potiuk potiuk reopened this Jul 1, 2026
@bugraoz93 bugraoz93 merged commit bfba4eb into apache:main Jul 1, 2026
575 of 587 checks passed
@github-actions

github-actions Bot commented Jul 1, 2026

Copy link
Copy Markdown
Contributor

Backport failed to create: v3-3-test. View the failure log Run details

Note: As of Merging PRs targeted for Airflow 3.X
the committer who merges the PR is responsible for backporting the PRs that are bug fixes (generally speaking) to the maintenance branches.

In matter of doubt please ask in #release-management Slack channel.

Status Branch Result
v3-3-test Commit Link

You can attempt to backport this manually by running:

cherry_picker bfba4eb v3-3-test

This should apply the commit to the v3-3-test branch and leave the commit in conflict state marking
the files that need manual conflict resolution.

After you have resolved the conflicts, you can continue the backport process by running:

cherry_picker --continue

If you don't have cherry-picker installed, see the installation guide.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

all versions If set, the CI build will be forced to use all versions of Python/K8S/DBs area:dev-tools backport-to-v3-3-test Backport to v3-3-test

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants