prole/infrastructure/roles/k3s/defaults/main.yml
chrisfu 2c6df07933 Harden monitoring init; improve k3s reset cleanup
- Add diagnostics/reset/retry logic for stuck monitoring Helm installs\n- Add optional cluster-side node/lease deletion during k3s reset via delegated control-plane\n- Bump prole-db image tag/version to 18-123\n- Update monitoring init unit test for reset/diagnostics path
2026-03-15 16:02:15 -07:00

74 lines
2.5 KiB
YAML

---
k3s_enabled: true
k3s_state: present
k3s_version: ""
k3s_guard_token_drift: true
# k3s reset workflow
# By default, cleanup/uninstall is local-only. The `k3s_reset` playbook enables
# cluster-side node removal explicitly.
k3s_reset_force_cluster_node_delete: false
k3s_reset_kubeconfig_path: /etc/rancher/k3s/k3s.yaml
# k3s service startup can be slow on first boot (image pulls, containerd unpacking, etc.).
# The service task is best-effort and then Ansible polls `systemctl is-active` using these knobs.
k3s_service_start_retries: 60
k3s_service_start_delay: 5
# Destructive repair/refresh workflow (server only)
# A refresh wipes the k3s installation/state to a blank slate but preserves the
# join token and CA material so existing agents can rejoin without changing the token.
k3s_refresh_backup_root: /root/k3s-refresh-backups
k3s_refresh_confirm: ""
k3s_refresh_run_configure: true
# CloudNative-PG
# Pin operator/plugin version by default to avoid unanticipated upgrades and load spikes.
# Upgrades must be explicitly enabled via `k3s_cnpg_upgrade: true`.
k3s_cnpg_version_default: "1.28.1"
k3s_cnpg_upgrade: false
k3s_role: agent
k3s_cluster_init: false
k3s_server_url: ""
k3s_token: "{{ vault_k3s_token | default('') }}"
# Optional: if `k3s_token` is a full node-token (`K10...::server:...`), set this
# to the raw secret (`...` after `::server:`) for the server config.
k3s_server_token: ""
k3s_disable: []
k3s_tls_sans: []
k3s_service_node_labels: []
k3s_node_taints: []
k3s_disable_agent: false
k3s_write_kubeconfig_mode: "0640"
k3s_kubeconfig_group: kubeadm
k3s_kubeconfig_users: []
k3s_registry_config_enabled: true
k3s_registry_config_path: /etc/rancher/k3s/registries.yaml
k3s_registry_host: ""
k3s_registry_namespace: ""
k3s_registry_port: 5000
# Pulling large images on every run can cause major load spikes; keep this opt-in.
k3s_prestage_images: false
k3s_prestage_images_platform: "linux/arm64"
k3s_prestage_images_list:
- docker.io/library/registry:2
- docker.io/prom/prometheus:v2.55.1
- docker.io/prom/alertmanager:v0.27.0
- quay.io/prometheus/node-exporter:v1.8.2
- quay.io/prometheus-operator/prometheus-operator:v0.71.2
- ghcr.io/openbao/openbao:2.0.0
- ghcr.io/opentofu/opentofu:1.8.2
- docker.io/library/kong:3.8
k3s_cert_manager_version: "v1.19.4"
k3s_svc_check_domain: "svc.prole.org"
k3s_svc_check_namespace: "svc-check"
k3s_svc_check_kong_namespace: "${SERVICE_NAMESPACE}"
k3s_svc_check_kong_configmap_name: "prole-svc-kong-config"
k3s_kong_namespace: "${SERVICE_NAMESPACE}"
k3s_kong_deployment: "prole-svc-kong"