菜单
Perform Talos node operations — apply a config change, add or change a system extension, upgrade Talos or Kubernetes, reboot or reset a node. Use for "apply talos config", "upgrade talos node", "reboot node", "add a talos extension", or "node config change". Scoped to the Artemis cluster.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/Exikle/Artemis-Cluster --skill talos-ops命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
Take a fast read on overall cluster health — failing pods, unready Flux resources, node and storage state — and report what is actually broken. Use for "cluster status", "what's broken", "health check", "anything down", or "quick status". Scoped to the Artemis cluster.
Find real-world HelmRelease examples for an app by searching kubesearch.dev and other home-ops repos, then adapt the best match to this cluster's conventions. Use for "find examples for X", "how do others deploy X", "search kubesearch for X", or when deploying an unfamiliar app with no obvious reference. Scoped to the Artemis cluster.
Watch an in-flight deployment or Flux reconciliation until it settles or fails, reporting state transitions as they happen. Use for "watch the deploy", "monitor the rollout", "keep an eye on flux", or when a change was just pushed and its landing needs following. Scoped to the Artemis cluster.
基于 SOC 职业分类
正在显示 SKILL.md
| name | talos-ops |
| description | Perform Talos node operations — apply a config change, add or change a system extension, upgrade Talos or Kubernetes, reboot or reset a node. Use for "apply talos config", "upgrade talos node", "reboot node", "add a talos extension", or "node config change". Scoped to the Artemis cluster. |
Node operations for Artemis-Cluster (Talos Linux with render-config / Jinja2 templates).
Also read
.agents/references/talos.mdfor node IPs and config background.
talos/nodes/<node>.yaml.j2just talos render-config <node>just talos apply-node <node>just talos render-config <node>
just talos apply-node <node> --mode=reboot
Drain the node first if it runs stateful workloads:
kubectl drain <node> --ignore-daemonsets --delete-emptydir-data
# apply + reboot
kubectl uncordon <node>
Extensions need a new schematic ID and reboot:
talos/schematics/<node>.yamljust talos gen-schematic-id <schematic> — registers with Sidero Image Factory, prints new IDtalos/nodes/<node>.yaml.j2 with the new schematic IDjust talos apply-node <node> --mode=reboottalosctl patch machineconfig with a full configFull config passed as a patch causes array fields (machine.files, machine.certSANs) to be duplicated in the merge, which can produce a read-only filesystem on next boot. Always use:
just talos apply-node <node> — for full config replacement (render-config output)Minimal certSANs patch example:
machine:
certSANs:
- <ip-or-hostname>
cluster:
apiServer:
certSANs:
- <ip-or-hostname>
Apply: talosctl apply-config --file patch.yaml -n <node-ip>
talosctl health -n <node-ip>
talosctl dmesg -n <node-ip> | tail -30
talosctl version -n <node-ip>
talosctl get machineconfig -n <node-ip> -o yaml
Check tuppr plans before manually upgrading — it manages automated Talos + Kubernetes upgrades:
kubectl get plans -n system-upgrade
kubectl get jobs -n system-upgrade -o wide
kubectl drain <node> --ignore-daemonsets --delete-emptydir-data
talosctl upgrade -n <node-ip> \
--image factory.talos.dev/installer/<schematic-id>:<version> \
--wait
kubectl uncordon <node>
Get the schematic ID from talos/schematics/<node>.yaml after running just talos gen-schematic-id.
| Node | Role | IP |
|---|---|---|
talos-cp-01 | Control plane | 10.10.99.x |
talos-cp-02 | Control plane | 10.10.99.x |
talos-cp-03 | Control plane | 10.10.99.x |
talos-w-01 | Worker | 10.10.99.x |
talos-w-02 | Worker | 10.10.99.x |
talos-gpu-01 | Worker + GPU (Arc A380) | 10.10.99.x |
Exact IPs: see
.agents/references/talos.mdorreference_ips.mdin memory.