Deploy k3s and Docker services via Ansible with locking, dry-run validation, and health verification.
npx skills add https://github.com/DanielH2018/server --skill deploy-danielh2018
Please help me install this Agent Skill. Skill: deploy Source: https://github.com/DanielH2018/server/tree/main/.claude/skills/deploy Command: npx skills add https://github.com/DanielH2018/server --skill deploy-danielh2018
ansible/roles/k8s/ or Docker roles on daniel-pi — and never runs Docker verification on cluster nodes where Docker no longer exists. - Three validation modes: Distinguishes prek (local schema checks), --check (task wiring only), and --dry-run (server-side validation against the live API server, including CRDs), so the right check answers the right question. - Locked, snapshot-based deploys: Runs through scripts/deploy.sh, which holds the shared git-tree lock, deploys from a snapshot of HEAD, and serializes per-service deploys; exit codes 2–77 distinguish "nothing deployed" resume points from real playbook failures. - Health verification gate: Confirms rollouts with probe.py health, which requires a fully complete rollout and no container restarts in the last 180 seconds — catching crashloops that kubectl rollout status reports as green. - Use Case: After merging a PR that changes a Traefik middleware, deploy the affected service with a server-side dry run first, then deploy and verify the rollout is healthy before declaring success. ## Quick Start Ask the assistant to deploy a specific service by name, for example requesting a dry run first and then a verified deploy of the service through the Ansible deploy wrapper.High-intent search queries and answers about installing and using this skill.
Run ./scripts/deploy.sh --tags "<service-name>" from /home/ubuntu/server. The wrapper takes the shared git-tree lock, snapshots HEAD into a worktree, and runs the playbook under a per-service lock. Add -e target=daniel-pi for Pi Docker services.
prek renders and schema-checks manifests locally without an API server. --check skips the apply entirely and only validates task wiring. --dry-run applies with kubectl --dry-run=server, so the live API server validates CRDs, ordering, and admission rules.
Exit 75 means a lock stayed busy and nothing was deployed — retry later. Exit 4 means the tree is behind origin/master; run git pull rather than bypassing the staleness check, since a stale tree renders stale templates.
No. Docker was uninstalled from daniel-box and daniel-server, so docker commands do not exist there. Verify k3s workloads with probe.py health <service>, which checks rollout completion and recent container restarts; use --docker only for the Pi.
Ansible reporting ok only means the playbook ran, not that the workload is healthy. A manifest can apply cleanly and then crash-loop or fail probes. Run probe.py health, then inspect with kubectl rollout status, describe pod, and logs.
Dry run refuses roles listed in k8s_dry_run_unsupported that mutate outside the manifests tree, and it cannot verify scheduling, PVC binding, probes, or rollout behavior. A brand-new service is only half-checked because volume-claim is skipped.
nq-rdlCreate, modify, and lint Ansible playbooks, roles, and inventories.
carter2099Tail k3s service logs by automating namespace and label lookups.
microscalerEncrypt and rotate deployment secrets with SOPS for GitOps repositories.
theslashdojoEncrypt and manage Ansible secrets in Git-backed playbooks using ansible-vault.