GKE Upgrade & Maintenance Advisor
Plans and validates upgrades and maintenance for Standard and Autopilot GKE clusters.
The material addresses maintenance exclusions, version skew, rollback, regional clusters, PDBs, GPU risks, and credential rotation, with some recovery guidance. However, it does not clearly require user confirmation for high-impact actions and gives limited data-flow, permission-boundary, or sensitive-data guidance. Points are deducted for incomplete least-privilege, confirmation, and pre/post-change safeguards.
The skill provides coherent preflight, upgrade, validation, rollback, and troubleshooting paths. Static review cannot reproduce the key paths, and several fixes directly recommend deleting bare pods or webhooks or relaxing PDBs without sufficient condition checks, backup confirmation, or controlled failure feedback. Under static calibration the score is capped at 10, with deductions for limited edge-case and failure-handling evidence.
Trigger terms, Standard versus Autopilot distinctions, version and channel inputs, workload concerns, and non-fit scenarios are clearly described. However, Chinese-language output and mainland-China network reachability are not addressed, while many procedures depend on gcloud, kubectl, the GKE control plane, and Google Cloud documentation. Points are deducted for unstated language, network, and environment prerequisites.
The skill has front matter, layered sections, linked templates, command examples, checklists, limitation disclosures, and Apache-2.0 licensing. The repository README indicates active development and provides installation and issue-reporting paths. The selected skill itself lacks a version, changelog, named maintenance owner, and explicit update process. Points are deducted for incomplete version and governance details.
The intended outputs are clear, and the templates and commands can be adapted into upgrade plans, checklists, and runbooks covering common workloads and failure modes. They still require environment-specific inputs and human validation, and some high-impact remediation advice requires substantial review. Under static calibration the score is capped at 7, with deductions for unverified correctness and limited direct usability evidence.
The skill links to official GKE release, upgrade, maintenance-exclusion, and rollout-sequencing references, and its commands, checklists, and diagnostic steps are auditable. There is no committed test suite, CI coverage, or third-party execution evidence, and this review did not independently reproduce the procedures. Under static calibration the score is capped at 5, with deductions for limited reproduction and corroboration.
- Commands that delete bare pods or admission webhooks, relax PDBs, or rotate credentials can cause service disruption or security impact; verify targets, back up configuration, and obtain user confirmation before execution.
- The documentation depends on Google Cloud/GKE services and externally hosted documentation, with no mainland-China reachability or offline alternative stated.
- GKE behavior, supported versions, and gcloud flags may change; verify the target environment against current version-specific official documentation and CLI help.
What does this skill do, and when should you use it?
This skill focuses on Google Kubernetes Engine cluster upgrades and maintenance for both Standard and Autopilot modes. It uses version, release-channel, topology, sequencing, and workload context to produce upgrade plans, checklists, runbooks, and troubleshooting guidance. Its scope includes control-plane and node-pool upgrades, maintenance windows and exclusions, release-channel strategy, PDB handling, stateful workloads, and GPU/TPU considerations. It is a fit for structured GKE lifecycle operations, not cluster creation, application onboarding, general networking, or security-policy configuration.
Before creating an upgrade artifact, it gathers cluster mode, current and target versions, release channel, environment topology, Rollout Sequencing usage, and workload sensitivity. It distinguishes minor and patch upgrades, plans sequential control-plane progression and node-pool ordering, and recommends Surge or Blue-Green strategies for Standard node pools. It produces runbooks containing gcloud and kubectl commands and copyable pre- and post-upgrade checklists. For failures, it examines PDB disruption limits, pending pods and capacity, bare pods, admission webhooks, and PVC attachment issues. For GPU and TPU environments, it addresses restart behavior, reservations, quota, driver and CUDA compatibility, and validation or rollback concerns.
- A platform team needs a staged upgrade plan across development, staging, and production GKE clusters.
- An operator needs to choose a release channel or configure maintenance windows and exclusions.
- A Standard cluster requires different upgrade settings for stateless, database, or GPU node pools.
- A team is investigating an upgrade blocked by PDBs, capacity, webhooks, bare pods, or PVC attachment.
- A GPU training platform needs to pause automatic upgrades and validate driver and CUDA compatibility before production maintenance.
What are this skill's strengths and limitations?
- Supports both Standard and Autopilot GKE clusters.
- Centers guidance on automatic upgrades, maintenance controls, release channels, and staged environments.
- Addresses stateful, GPU, batch, and operator-related workload concerns.
- Produces runbooks with gcloud and kubectl commands plus upgrade checklists.
- Systematically covers five major categories of stuck-upgrade causes.
- Useful output depends on cluster, version, channel, topology, and workload context; vague inputs require assumptions.
- The complete runbook, checklist, and diagnostic templates are referenced but not included in the supplied source.
- No test suite, execution evidence, or platform validation record is provided.
- Actual upgrades still depend on GKE, gcloud, kubectl, and environment capacity, quota, and permissions.
How do you install this skill?
The skill is located at skills/cloud/gke-upgrades/SKILL.md. To install the Google Skills collection, run npx skills add google/skills, then select the desired skill from the repository. The source does not document a separate installation command or a specific client directory layout for this skill.
How do you use this skill?
After placing the skill in an Agent Skills-compatible client, describe a concrete task in natural language, such as “plan our next GKE maintenance” or “our GKE upgrade is stuck.” Supplying cluster mode, current and target versions, release channel, topology, and workload types enables a more tailored result. The source does not document a fixed command entry point or provide the full contents of the referenced templates.
How does this skill compare with similar options?
The skill explicitly excludes cluster creation, application onboarding, general networking, and security-policy configuration, directing those needs to gke-basics or other relevant GKE skills.