kueue-release-notes

Skill: Kueue Release Notes

When to use this skill

Use this skill when writing or reviewing a Kueue release note for a pull request.

The goal is to produce a concise, professional release note describing the user-observable change rather than only the implementation detail.

Output

Final output is three release-note variants plus rationale for key decisions:

• Detailed: one or two sentences with enough context to explain the user-visible scenario.
• Concise: the shortest accurate release-note bullet.
• Balanced: the recommended version, balancing clarity and brevity.
• Rationale: briefly explain the prefix, framing, user-observable impact, and any formatting or mitigation decisions.

If the user asks for only one version, provide the Balanced variant.

Critical rule for breaking changes

For breaking changes, include a mitigation section starting exactly with ACTION REQUIRED:.

Use this when users must change manifests, Helm values, CLI flags, configuration, workloads, namespace layout, RBAC, CRDs, or operational procedures before or during upgrade.

The mitigation must be specific and actionable. For installer-specific behavior, distinguish Helm and manifest-based installs explicitly.

Example:

<Prefix>: <short description of changed behavior>.

ACTION REQUIRED: If <affected scenario>, <mitigation before upgrading>. <Optional caveat>.

Process

1. Read the PR and classify the change

Read the PR description, linked issue, release-note text, and relevant changed files.

Identify:

• the affected Kueue area or component;
• whether this is a bugfix, feature, observability improvement, documentation-only change, cleanup, or breaking change;
• the scenario users could observe before the change;
• the behavior users should expect after the change;
• whether installation mode matters, for example Helm versus manifest-based installs;
• whether users need to take action before upgrading.

Do not rely only on the PR title. PR titles often describe the implementation, while release notes should describe the user-visible effect.

2. Apply general principles

Focus on user-observable behavior

Describe what changes for users, operators, or integrators.

Good user-observable aspects include:

• workloads being admitted, blocked, evicted, requeued, or left inactive;
• queueing, quota, scheduling, or preemption behavior;
• webhook behavior;
• generated manifests or Helm chart behavior;
• API fields, feature gates, flags, configuration, and CRDs;
• logs, events, metrics, and visibility endpoints, because these are part of the observability surface;
• UI behavior in KueueViz.

Avoid release notes that only say what code changed, such as:

• “refactored locking”;
• “added helper function”;
• “changed timeout implementation”;
• “updated tests”;
• “renamed internal variable”.

If an implementation detail is important to understand the user-visible behavior, include it briefly after the scenario.

Keep it concise and professional

Default to one bullet with one or two sentences.

Use:

• past tense for fixes: Fixed, Improved, Removed;
• active voice;
• code formatting for fields, flags, metrics, feature gates, resource names, and configuration keys;
• precise nouns such as Workload, ClusterQueue, LocalQueue, MultiKueueCluster, AdmissionCheck, namespaceSelector.

Avoid:

• vague correctness claims such as “fixed handling” without describing the expected behavior;
• “various” or “some”;
• excessive implementation details;
• blame-oriented wording;
• speculative impact;
• very long lists of changed files or internal functions.

3. Apply type-specific framing

Bugfix

For bugfixes, make the broken scenario clear.

Preferred shape:

<Prefix>: Fixed a bug where <scenario> could <bad user-visible effect>.

Optional second sentence:

Kueue now <brief fix or new behavior>.

Examples of good bugfix framing:

MultiKueue: Fixed a bug where one slow or unresponsive remote cluster could stall reconciliation for other MultiKueueClusters, even when `controller.groupKindConcurrency["MultiKueueCluster.kueue.x-k8s.io"]` was set above 1. This could delay or block workload admission through other healthy clusters.

KueueViz: Fixed the navigation bar to avoid layout breakage on narrow mobile screens.

Avoid starting with implementation details:

MultiKueue: Changed remote watch establishment to use per-cluster locking.

That implementation detail can be included only if it explains the user-visible outcome.

Feature or enhancement

For features and enhancements, describe the new capability and why users or operators benefit from it.

Preferred shape:

<Prefix>: Added <capability>, allowing users to <user-visible outcome>.

If the feature is gated, mention the feature gate or configuration field only when users need it to use the capability.

Observability

Logs, metrics, events, and visibility endpoints are user-facing for operators.

Use Observability: when the main effect is improved debugging, reduced noise, corrected metrics, or clearer operational signals.

Examples:

Observability: Fixed `kueue_cohort_subtree_quota` and `kueue_cohort_subtree_resource_reservations` to report CPU resources in CPU units instead of raw milliCPU values.

Observability: Improved FairSharing strategy-evaluation logs by including DRS share values at verbosity level V(4).

Breaking change

Follow the critical breaking-change rule near the top of this skill.

The mitigation must tell affected users exactly what to change before upgrading.

When behavior differs by installation method, explicitly distinguish manifest-based and Helm-based installations.

Example mitigation details may include:

• which field, flag, Helm value, or manifest entry to change;
• whether the change applies only to customized installations;
• whether a workaround is discouraged;
• what behavior users should expect after applying the mitigation.

Removal of deprecated option

For deprecated option removals, name the removed option and its replacement.

Preferred shape:

<Prefix>: Removed the deprecated `<flag-or-field>` option. Use `<replacement>` instead.

Include a breaking-change mitigation when users must update their configuration before upgrading.

4. Choose a narrow prefix

Start the note with a prefix that narrows the scope.

Prefer the smallest accurate user-facing area, for example:

• MultiKueue:
• TAS:
• Helm:
• KueueViz:
• Observability:
• Scheduling:
• VisibilityOnDemand:
• LeaderWorkerSet:
• ElasticJobsViaWorkloadSlices:
• <integration name>: such as JobSet:, StatefulSet:, or RayJob:

Use a combined prefix when the change is intentionally scoped to two areas, for example:

• LeaderWorkerSet & StatefulSet:
• Helm and manifests:

Avoid broad prefixes such as Kueue: unless the change truly affects the whole project or no narrower area applies.

5. Validate accuracy before finalizing

Before proposing the final note, use this checklist as a starting point and add any PR-specific checks needed for accuracy:

• Does the prefix describe the narrowest accurate scope?
• Does the note describe a user-visible change?
• For a bugfix, does it name the broken scenario?
• For observability, does it mention the log, metric, event, or endpoint users see?
• For breaking changes, does it include a specific mitigation?
• Are Helm and manifest-based installations distinguished when relevant?
• Are API fields, flags, metrics, and feature gates spelled exactly as in the code?
• Is the note concise enough to fit as a release bullet?

6. Propose the variants and rationale

Provide the final answer in this order:

1. Detailed
2. Concise
3. Balanced
4. Rationale

The rationale should expand on the key decisions, especially:

• why the prefix is the narrowest accurate scope;
• why the change is framed as a bugfix, feature, observability improvement, breaking change, cleanup, or documentation-only change;
• what user-observable behavior the note highlights;
• whether mitigation is needed;
• why the recommended version is the best balance of clarity and brevity.

Templates

Bugfix

<Prefix>: Fixed a bug where <scenario> could <user-visible impact>.

Bugfix with brief implementation note

<Prefix>: Fixed a bug where <scenario> could <user-visible impact>. Kueue now <brief new behavior>.

Feature or enhancement

<Prefix>: Added <capability>, allowing users to <user-visible outcome>.

Observability

Observability: Improved <logs|metrics|events|visibility endpoint> by <specific change>, making it easier to <debug|monitor|understand user-visible behavior>.

Breaking change

<Prefix>: <short description of changed behavior>.

ACTION REQUIRED: If <affected scenario>, <mitigation before upgrading>. <Optional caveat>.

Removal of deprecated option

<Prefix>: Removed the deprecated `<flag-or-field>` option. Use `<replacement>` instead.

ACTION REQUIRED: If your installation uses `<flag-or-field>`, remove it before upgrading and configure `<replacement>`.