111

deployment/ansible/playbooks/configure/configure-nomad-dynamic-volumes.yml

@@ -0,0 +1,62 @@
+---
+- name: Configure Nomad Dynamic Host Volumes for NFS
+  hosts: nomad_clients
+  become: yes
+  vars:
+    nfs_server: "snail"
+    nfs_share: "/fs/1000/nfs/Fnsync"
+    mount_point: "/mnt/fnsync"
+
+  tasks:
+    - name: Stop Nomad service
+      systemd:
+        name: nomad
+        state: stopped
+
+    - name: Update Nomad configuration for dynamic host volumes
+      blockinfile:
+        path: /etc/nomad.d/nomad.hcl
+        marker: "# {mark} DYNAMIC HOST VOLUMES CONFIGURATION"
+        block: |
+          client {
+            # 启用动态host volumes
+            host_volume "fnsync" {
+              path = "{{ mount_point }}"
+              read_only = false
+            }
+            
+            # 添加NFS相关的节点元数据
+            meta {
+              nfs_server = "{{ nfs_server }}"
+              nfs_share = "{{ nfs_share }}"
+              nfs_mounted = "true"
+            }
+          }
+        insertafter: 'client {'
+
+    - name: Start Nomad service
+      systemd:
+        name: nomad
+        state: started
+        enabled: yes
+
+    - name: Wait for Nomad to start
+      wait_for:
+        port: 4646
+        delay: 10
+        timeout: 60
+
+    - name: Check Nomad status
+      command: nomad node status
+      register: nomad_status
+      ignore_errors: yes
+
+    - name: Display Nomad status
+      debug:
+        var: nomad_status.stdout_lines
+
+
+
+
+
+

deployment/ansible/playbooks/deploy-nomad-config.yml

@@ -0,0 +1,41 @@
+---
+- name: 部署Nomad服务器配置模板
+  hosts: nomad_servers
+  become: yes
+  
+  tasks:
+    - name: 部署Nomad配置文件
+      template:
+        src: nomad-server.hcl.j2
+        dest: /etc/nomad.d/nomad.hcl
+        backup: yes
+        owner: root
+        group: root
+        mode: '0644'
+    
+    - name: 重启Nomad服务
+      systemd:
+        name: nomad
+        state: restarted
+        enabled: yes
+    
+    - name: 等待Nomad服务启动
+      wait_for:
+        port: 4646
+        host: "{{ ansible_host }}"
+        timeout: 30
+    
+    - name: 显示Nomad服务状态
+      systemd:
+        name: nomad
+      register: nomad_status
+    
+    - name: 显示服务状态
+      debug:
+        msg: "{{ inventory_hostname }} Nomad服务状态: {{ nomad_status.status.ActiveState }}"
+
+
+
+
+
+

deployment/ansible/playbooks/fix-bootstrap-expect.yml

@@ -0,0 +1,39 @@
+---
+- name: 紧急修复Nomad bootstrap_expect配置
+  hosts: nomad_servers
+  become: yes
+  
+  tasks:
+    - name: 修复bootstrap_expect为3
+      lineinfile:
+        path: /etc/nomad.d/nomad.hcl
+        regexp: '^  bootstrap_expect = \d+'
+        line: '  bootstrap_expect = 3'
+        backup: yes
+    
+    - name: 重启Nomad服务
+      systemd:
+        name: nomad
+        state: restarted
+        enabled: yes
+    
+    - name: 等待Nomad服务启动
+      wait_for:
+        port: 4646
+        host: "{{ ansible_host }}"
+        timeout: 30
+    
+    - name: 检查Nomad服务状态
+      systemd:
+        name: nomad
+      register: nomad_status
+    
+    - name: 显示Nomad服务状态
+      debug:
+        msg: "{{ inventory_hostname }} Nomad服务状态: {{ nomad_status.status.ActiveState }}"
+
+
+
+
+
+

deployment/ansible/playbooks/fix-ch4-nomad-config.yml

@@ -0,0 +1,103 @@
+---
+- name: Fix ch4 Nomad configuration - convert from server to client
+  hosts: ch4
+  become: yes
+  vars:
+    ansible_host: 100.117.106.136
+  
+  tasks:
+    - name: Backup current Nomad config
+      copy:
+        src: /etc/nomad.d/nomad.hcl
+        dest: /etc/nomad.d/nomad.hcl.backup
+        remote_src: yes
+        backup: yes
+
+    - name: Update Nomad config to client mode
+      blockinfile:
+        path: /etc/nomad.d/nomad.hcl
+        marker: "# {mark} ANSIBLE MANAGED CLIENT CONFIG"
+        block: |
+          server {
+            enabled = false
+          }
+
+          client {
+            enabled = true
+            network_interface = "tailscale0"
+            
+            servers = [
+              "semaphore.tailnet-68f9.ts.net:4647",
+              "ash1d.tailnet-68f9.ts.net:4647",
+              "ash2e.tailnet-68f9.ts.net:4647",
+              "ch2.tailnet-68f9.ts.net:4647",
+              "ch3.tailnet-68f9.ts.net:4647",
+              "onecloud1.tailnet-68f9.ts.net:4647",
+              "de.tailnet-68f9.ts.net:4647"
+            ]
+            
+            meta {
+              consul = "true"
+              consul_version = "1.21.5"
+              consul_server = "true"
+            }
+          }
+        insertbefore: '^server \{'
+        replace: '^server \{.*?\}'
+
+    - name: Update client block
+      blockinfile:
+        path: /etc/nomad.d/nomad.hcl
+        marker: "# {mark} ANSIBLE MANAGED CLIENT BLOCK"
+        block: |
+          client {
+            enabled = true
+            network_interface = "tailscale0"
+            
+            servers = [
+              "semaphore.tailnet-68f9.ts.net:4647",
+              "ash1d.tailnet-68f9.ts.net:4647",
+              "ash2e.tailnet-68f9.ts.net:4647",
+              "ch2.tailnet-68f9.ts.net:4647",
+              "ch3.tailnet-68f9.ts.net:4647",
+              "onecloud1.tailnet-68f9.ts.net:4647",
+              "de.tailnet-68f9.ts.net:4647"
+            ]
+            
+            meta {
+              consul = "true"
+              consul_version = "1.21.5"
+              consul_server = "true"
+            }
+          }
+        insertbefore: '^client \{'
+        replace: '^client \{.*?\}'
+
+    - name: Restart Nomad service
+      systemd:
+        name: nomad
+        state: restarted
+        enabled: yes
+
+    - name: Wait for Nomad to be ready
+      wait_for:
+        port: 4646
+        host: "{{ ansible_default_ipv4.address }}"
+        delay: 5
+        timeout: 30
+
+    - name: Verify Nomad client status
+      shell: |
+        NOMAD_ADDR=http://localhost:4646 nomad node status | grep -q "ready"
+      register: nomad_ready
+      failed_when: nomad_ready.rc != 0
+      retries: 3
+      delay: 10
+
+    - name: Display completion message
+      debug:
+        msg: |
+          ✅ Successfully converted ch4 from Nomad server to client
+          ✅ Nomad service restarted
+          ✅ Configuration updated
+

deployment/ansible/playbooks/fix-master-to-ch4.yml

@@ -0,0 +1,82 @@
+---
+- name: Fix master node - rename to ch4 and restore SSH port 22
+  hosts: master
+  become: yes
+  vars:
+    new_hostname: ch4
+    old_hostname: master
+  
+  tasks:
+    - name: Backup current hostname
+      copy:
+        content: "{{ old_hostname }}"
+        dest: /etc/hostname.backup
+        mode: '0644'
+      when: ansible_hostname == old_hostname
+
+    - name: Update hostname to ch4
+      hostname:
+        name: "{{ new_hostname }}"
+      when: ansible_hostname == old_hostname
+
+    - name: Update /etc/hostname file
+      copy:
+        content: "{{ new_hostname }}"
+        dest: /etc/hostname
+        mode: '0644'
+      when: ansible_hostname == old_hostname
+
+    - name: Update /etc/hosts file
+      lineinfile:
+        path: /etc/hosts
+        regexp: '^127\.0\.1\.1.*{{ old_hostname }}'
+        line: '127.0.1.1 {{ new_hostname }}'
+        state: present
+      when: ansible_hostname == old_hostname
+
+    - name: Update Tailscale hostname
+      shell: |
+        tailscale set --hostname={{ new_hostname }}
+      when: ansible_hostname == old_hostname
+
+    - name: Backup SSH config
+      copy:
+        src: /etc/ssh/sshd_config
+        dest: /etc/ssh/sshd_config.backup
+        remote_src: yes
+        backup: yes
+
+    - name: Restore SSH port to 22
+      lineinfile:
+        path: /etc/ssh/sshd_config
+        regexp: '^Port '
+        line: 'Port 22'
+        state: present
+
+    - name: Restart SSH service
+      systemd:
+        name: ssh
+        state: restarted
+        enabled: yes
+
+    - name: Wait for SSH to be ready on port 22
+      wait_for:
+        port: 22
+        host: "{{ ansible_default_ipv4.address }}"
+        delay: 5
+        timeout: 30
+
+    - name: Test SSH connection on port 22
+      ping:
+      delegate_to: "{{ inventory_hostname }}"
+      vars:
+        ansible_port: 22
+
+    - name: Display completion message
+      debug:
+        msg: |
+          ✅ Successfully renamed {{ old_hostname }} to {{ new_hostname }}
+          ✅ SSH port restored to 22
+          ✅ Tailscale hostname updated
+          🔄 Please update your inventory file to use the new hostname and port
+

deployment/ansible/playbooks/install-consul-clients.yml

@@ -0,0 +1,71 @@
+---
+- name: Install and configure Consul clients on all nodes
+  hosts: all
+  become: yes
+  vars:
+    consul_servers:
+      - "100.117.106.136"  # ch4 (韩国)
+      - "100.122.197.112"  # warden (北京)
+      - "100.116.80.94"    # ash3c (美国)
+  
+  tasks:
+    - name: Get Tailscale IP address
+      shell: ip addr show tailscale0 | grep 'inet ' | awk '{print $2}' | cut -d/ -f1
+      register: tailscale_ip_result
+      changed_when: false
+
+    - name: Set Tailscale IP fact
+      set_fact:
+        tailscale_ip: "{{ tailscale_ip_result.stdout }}"
+
+    - name: Install Consul
+      apt:
+        name: consul
+        state: present
+        update_cache: yes
+
+    - name: Create Consul data directory
+      file:
+        path: /opt/consul/data
+        state: directory
+        owner: consul
+        group: consul
+        mode: '0755'
+
+    - name: Create Consul log directory
+      file:
+        path: /var/log/consul
+        state: directory
+        owner: consul
+        group: consul
+        mode: '0755'
+
+    - name: Create Consul config directory
+      file:
+        path: /etc/consul.d
+        state: directory
+        owner: consul
+        group: consul
+        mode: '0755'
+
+    - name: Generate Consul client configuration
+      template:
+        src: consul-client.hcl.j2
+        dest: /etc/consul.d/consul.hcl
+        owner: consul
+        group: consul
+        mode: '0644'
+      notify: restart consul
+
+    - name: Enable and start Consul service
+      systemd:
+        name: consul
+        enabled: yes
+        state: started
+        daemon_reload: yes
+
+  handlers:
+    - name: restart consul
+      systemd:
+        name: consul
+        state: restarted

deployment/ansible/playbooks/install/install-nfs-csi-plugin.yml

@@ -0,0 +1,91 @@
+---
+- name: Install NFS CSI Plugin for Nomad
+  hosts: nomad_nodes
+  become: yes
+  vars:
+    nomad_user: nomad
+    nomad_plugins_dir: /opt/nomad/plugins
+    csi_driver_version: "v4.0.0"
+    csi_driver_url: "https://github.com/kubernetes-csi/csi-driver-nfs/releases/download/{{ csi_driver_version }}/csi-nfs-driver"
+
+  tasks:
+    - name: Stop Nomad service
+      systemd:
+        name: nomad
+        state: stopped
+
+    - name: Create plugins directory
+      file:
+        path: "{{ nomad_plugins_dir }}"
+        state: directory
+        owner: "{{ nomad_user }}"
+        group: "{{ nomad_user }}"
+        mode: '0755'
+
+    - name: Download NFS CSI driver
+      get_url:
+        url: "{{ csi_driver_url }}"
+        dest: "{{ nomad_plugins_dir }}/csi-nfs-driver"
+        owner: "{{ nomad_user }}"
+        group: "{{ nomad_user }}"
+        mode: '0755'
+
+    - name: Install required packages for CSI
+      package:
+        name: 
+          - nfs-common
+          - mount
+        state: present
+
+    - name: Create CSI mount directory
+      file:
+        path: /opt/nomad/csi
+        state: directory
+        owner: "{{ nomad_user }}"
+        group: "{{ nomad_user }}"
+        mode: '0755'
+
+    - name: Update Nomad configuration for CSI plugin
+      blockinfile:
+        path: /etc/nomad.d/nomad.hcl
+        marker: "# {mark} CSI PLUGIN CONFIGURATION"
+        block: |
+          plugin_dir = "{{ nomad_plugins_dir }}"
+
+          plugin "csi-nfs" {
+            type = "csi"
+            config {
+              driver_name = "nfs.csi.k8s.io"
+              mount_dir = "/opt/nomad/csi"
+              health_timeout = "30s"
+              log_level = "INFO"
+            }
+          }
+        insertafter: 'data_dir = "/opt/nomad/data"'
+
+    - name: Start Nomad service
+      systemd:
+        name: nomad
+        state: started
+        enabled: yes
+
+    - name: Wait for Nomad to start
+      wait_for:
+        port: 4646
+        delay: 10
+        timeout: 60
+
+    - name: Check Nomad status
+      command: nomad node status
+      register: nomad_status
+      ignore_errors: yes
+
+    - name: Display Nomad status
+      debug:
+        var: nomad_status.stdout_lines
+
+
+
+
+
+

deployment/ansible/playbooks/start-nomad-servers.yml

@@ -0,0 +1,33 @@
+---
+- name: 启动所有Nomad服务器形成集群
+  hosts: nomad_servers
+  become: yes
+  
+  tasks:
+    - name: 检查Nomad服务状态
+      systemd:
+        name: nomad
+      register: nomad_status
+    
+    - name: 启动Nomad服务（如果未运行）
+      systemd:
+        name: nomad
+        state: started
+        enabled: yes
+      when: nomad_status.status.ActiveState != "active"
+    
+    - name: 等待Nomad服务启动
+      wait_for:
+        port: 4646
+        host: "{{ ansible_host }}"
+        timeout: 30
+    
+    - name: 显示Nomad服务状态
+      debug:
+        msg: "{{ inventory_hostname }} Nomad服务状态: {{ nomad_status.status.ActiveState }}"
+
+
+
+
+
+

deployment/ansible/playbooks/templates/consul-client.hcl.j2

@@ -0,0 +1,61 @@
+# Consul Client Configuration for {{ inventory_hostname }}
+datacenter = "dc1"
+data_dir = "/opt/consul/data"
+log_level = "INFO"
+node_name = "{{ inventory_hostname }}"
+bind_addr = "{{ hostvars[inventory_hostname]['tailscale_ip'] }}"
+
+# Client mode (not server)
+server = false
+
+# Connect to Consul servers (指向三节点集群)
+retry_join = [
+{% for server in consul_servers %}
+  "{{ server }}"{% if not loop.last %},{% endif %}
+{% endfor %}
+]
+
+# Performance optimization
+performance {
+  raft_multiplier = 5
+}
+
+# Ports configuration
+ports {
+  grpc = 8502
+  http = 8500
+  dns = 8600
+}
+
+# Enable Connect for service mesh
+connect {
+  enabled = true
+}
+
+# Cache configuration for performance
+cache {
+  entry_fetch_max_burst = 42
+  entry_fetch_rate = 30
+}
+
+# Node metadata
+node_meta = {
+  region = "unknown"
+  zone = "nomad-{{ 'server' if 'server' in group_names else 'client' }}"
+}
+
+# UI disabled for clients
+ui_config {
+  enabled = false
+}
+
+# ACL configuration (if needed)
+acl = {
+  enabled = false
+  default_policy = "allow"
+}
+
+# Logging
+log_file = "/var/log/consul/consul.log"
+log_rotate_duration = "24h"
+log_rotate_max_files = 7

deployment/ansible/playbooks/templates/nomad-server.hcl.j2

@@ -0,0 +1,106 @@
+datacenter = "dc1"
+data_dir = "/opt/nomad/data"
+plugin_dir = "/opt/nomad/plugins"
+log_level = "INFO"
+name = "{{ ansible_hostname }}"
+
+bind_addr = "0.0.0.0"
+
+addresses {
+  http = "{{ ansible_host }}"
+  rpc  = "{{ ansible_host }}"
+  serf = "{{ ansible_host }}"
+}
+
+advertise {
+  http = "{{ ansible_host }}:4646"
+  rpc  = "{{ ansible_host }}:4647"
+  serf = "{{ ansible_host }}:4648"
+}
+
+ports {
+  http = 4646
+  rpc  = 4647
+  serf = 4648
+}
+
+server {
+  enabled = true
+  bootstrap_expect = 3
+  server_join {
+    retry_join = [
+      "semaphore.tailnet-68f9.ts.net:4648",
+      "ash1d.tailnet-68f9.ts.net:4648", 
+      "ash2e.tailnet-68f9.ts.net:4648",
+      "ch2.tailnet-68f9.ts.net:4648",
+      "ch3.tailnet-68f9.ts.net:4648",
+      "onecloud1.tailnet-68f9.ts.net:4648",
+      "de.tailnet-68f9.ts.net:4648",
+      "hcp1.tailnet-68f9.ts.net:4648"
+    ]
+  }
+}
+
+{% if ansible_hostname == 'hcp1' %}
+client {
+  enabled = true
+  network_interface = "tailscale0"
+  
+  servers = [
+    "semaphore.tailnet-68f9.ts.net:4647",
+    "ash1d.tailnet-68f9.ts.net:4647",
+    "ash2e.tailnet-68f9.ts.net:4647",
+    "ch2.tailnet-68f9.ts.net:4647",
+    "ch3.tailnet-68f9.ts.net:4647",
+    "onecloud1.tailnet-68f9.ts.net:4647",
+    "de.tailnet-68f9.ts.net:4647",
+    "hcp1.tailnet-68f9.ts.net:4647"
+  ]
+  
+  host_volume "traefik-certs" {
+    path      = "/opt/traefik/certs"
+    read_only = false
+  }
+  
+  host_volume "fnsync" {
+    path = "/mnt/fnsync"
+    read_only = false
+  }
+  
+  meta {
+    consul = "true"
+    consul_version = "1.21.5"
+    consul_client = "true"
+  }
+  
+  gc_interval = "5m"
+  gc_disk_usage_threshold = 80
+  gc_inode_usage_threshold = 70
+}
+
+plugin "nomad-driver-podman" {
+  config {
+    socket_path = "unix:///run/podman/podman.sock"
+    volumes {
+      enabled = true
+    }
+  }
+}
+{% endif %}
+
+consul {
+  address = "ch4.tailnet-68f9.ts.net:8500,ash3c.tailnet-68f9.ts.net:8500,warden.tailnet-68f9.ts.net:8500"
+  server_service_name = "nomad"
+  client_service_name = "nomad-client"
+  auto_advertise = true
+  server_auto_join = false
+  client_auto_join = true
+}
+
+telemetry {
+  collection_interval = "1s"
+  disable_hostname = false
+  prometheus_metrics = true
+  publish_allocation_metrics = true
+  publish_node_metrics = true
+}

deployment/ansible/playbooks/update-hosts-file.yml

@@ -19,7 +19,7 @@
       - ip: "100.120.225.29"
         hostnames: ["de"]
       - ip: "100.117.106.136"
-        hostnames: ["master"]
+        hostnames: ["ch4"]
       - ip: "100.116.80.94"
         hostnames: ["ash3c", "influxdb1"]
       - ip: "100.116.112.45"

deployment/ansible/playbooks/update-nomad-peers.yml

@@ -0,0 +1,56 @@
+---
+- name: 更新Nomad服务器配置，添加hcp1作为peer
+  hosts: nomad_servers
+  become: yes
+  vars:
+    hcp1_ip: "100.97.62.111"
+    bootstrap_expect: 8
+  
+  tasks:
+    - name: 备份原配置文件
+      copy:
+        src: /etc/nomad.d/nomad.hcl
+        dest: /etc/nomad.d/nomad.hcl.bak
+        remote_src: yes
+        backup: yes
+    
+    - name: 添加hcp1到retry_join列表
+      lineinfile:
+        path: /etc/nomad.d/nomad.hcl
+        regexp: '^  retry_join = \['
+        line: '  retry_join = ["{{ hcp1_ip }}",'
+        backup: yes
+    
+    - name: 更新bootstrap_expect为8
+      lineinfile:
+        path: /etc/nomad.d/nomad.hcl
+        regexp: '^  bootstrap_expect = \d+'
+        line: '  bootstrap_expect = {{ bootstrap_expect }}'
+        backup: yes
+    
+    - name: 重启Nomad服务
+      systemd:
+        name: nomad
+        state: restarted
+        enabled: yes
+    
+    - name: 等待Nomad服务启动
+      wait_for:
+        port: 4646
+        host: "{{ ansible_host }}"
+        timeout: 30
+    
+    - name: 检查Nomad服务状态
+      systemd:
+        name: nomad
+      register: nomad_status
+    
+    - name: 显示Nomad服务状态
+      debug:
+        msg: "Nomad服务状态: {{ nomad_status.status.ActiveState }}"
+
+
+
+
+
+