Clean up repository: remove backup files and reorganize infrastructure components
This commit is contained in:
@@ -4,17 +4,6 @@
|
||||
become: yes
|
||||
vars:
|
||||
nomad_config_dir: /etc/nomad.d
|
||||
client_ip: "{{ ansible_host }}"
|
||||
|
||||
# Nomad节点名称（带地理位置前缀）
|
||||
client_name: >-
|
||||
{%- if inventory_hostname == 'influxdb1' -%}us-influxdb
|
||||
{%- elif inventory_hostname == 'master' -%}kr-master
|
||||
{%- elif inventory_hostname == 'hcp1' -%}bj-hcp1
|
||||
{%- elif inventory_hostname == 'hcp2' -%}bj-hcp2
|
||||
{%- elif inventory_hostname == 'warden' -%}bj-warden
|
||||
{%- else -%}{{ inventory_hostname }}
|
||||
{%- endif -%}
|
||||
|
||||
tasks:
|
||||
- name: 创建Nomad配置目录
|
||||
|
||||
@@ -1,104 +0,0 @@
|
||||
---
|
||||
- name: 配置Nomad客户端节点
|
||||
hosts: target_nodes
|
||||
become: yes
|
||||
vars:
|
||||
nomad_config_dir: /etc/nomad.d
|
||||
|
||||
tasks:
|
||||
- name: 创建Nomad配置目录
|
||||
file:
|
||||
path: "{{ nomad_config_dir }}"
|
||||
state: directory
|
||||
owner: root
|
||||
group: root
|
||||
mode: '0755'
|
||||
|
||||
- name: 复制Nomad客户端配置
|
||||
copy:
|
||||
content: |
|
||||
datacenter = "dc1"
|
||||
data_dir = "/opt/nomad/data"
|
||||
log_level = "INFO"
|
||||
bind_addr = "0.0.0.0"
|
||||
|
||||
server {
|
||||
enabled = false
|
||||
}
|
||||
|
||||
client {
|
||||
enabled = true
|
||||
# 配置七姐妹服务器地址
|
||||
servers = [
|
||||
"100.116.158.95:4647", # bj-semaphore
|
||||
"100.81.26.3:4647", # ash1d
|
||||
"100.103.147.94:4647", # ash2e
|
||||
"100.90.159.68:4647", # ch2
|
||||
"100.86.141.112:4647", # ch3
|
||||
"100.98.209.50:4647", # bj-onecloud1
|
||||
"100.120.225.29:4647" # de
|
||||
]
|
||||
host_volume "fnsync" {
|
||||
path = "/mnt/fnsync"
|
||||
read_only = false
|
||||
}
|
||||
# 禁用Docker驱动，只使用Podman
|
||||
options {
|
||||
"driver.raw_exec.enable" = "1"
|
||||
"driver.exec.enable" = "1"
|
||||
}
|
||||
}
|
||||
|
||||
# 配置Podman插件目录
|
||||
plugin_dir = "/opt/nomad/plugins"
|
||||
|
||||
addresses {
|
||||
http = "{{ ansible_host }}"
|
||||
rpc = "{{ ansible_host }}"
|
||||
serf = "{{ ansible_host }}"
|
||||
}
|
||||
|
||||
advertise {
|
||||
http = "{{ ansible_host }}:4646"
|
||||
rpc = "{{ ansible_host }}:4647"
|
||||
serf = "{{ ansible_host }}:4648"
|
||||
}
|
||||
|
||||
consul {
|
||||
address = "100.116.158.95:8500"
|
||||
}
|
||||
|
||||
# 配置Podman驱动
|
||||
plugin "podman" {
|
||||
config {
|
||||
volumes {
|
||||
enabled = true
|
||||
}
|
||||
logging {
|
||||
type = "journald"
|
||||
}
|
||||
gc {
|
||||
container = true
|
||||
}
|
||||
}
|
||||
}
|
||||
dest: "{{ nomad_config_dir }}/nomad.hcl"
|
||||
owner: root
|
||||
group: root
|
||||
mode: '0644'
|
||||
|
||||
- name: 启动Nomad服务
|
||||
systemd:
|
||||
name: nomad
|
||||
state: restarted
|
||||
enabled: yes
|
||||
daemon_reload: yes
|
||||
|
||||
- name: 检查Nomad服务状态
|
||||
command: systemctl status nomad
|
||||
register: nomad_status
|
||||
changed_when: false
|
||||
|
||||
- name: 显示Nomad服务状态
|
||||
debug:
|
||||
var: nomad_status.stdout_lines
|
||||
@@ -1,104 +0,0 @@
|
||||
---
|
||||
- name: 配置Nomad客户端节点
|
||||
hosts: target_nodes
|
||||
become: yes
|
||||
vars:
|
||||
nomad_config_dir: /etc/nomad.d
|
||||
|
||||
tasks:
|
||||
- name: 创建Nomad配置目录
|
||||
file:
|
||||
path: "{{ nomad_config_dir }}"
|
||||
state: directory
|
||||
owner: root
|
||||
group: root
|
||||
mode: '0755'
|
||||
|
||||
- name: 复制Nomad客户端配置
|
||||
copy:
|
||||
content: |
|
||||
datacenter = "dc1"
|
||||
data_dir = "/opt/nomad/data"
|
||||
log_level = "INFO"
|
||||
bind_addr = "0.0.0.0"
|
||||
|
||||
server {
|
||||
enabled = false
|
||||
}
|
||||
|
||||
client {
|
||||
enabled = true
|
||||
# 配置七姐妹服务器地址
|
||||
servers = [
|
||||
"100.116.158.95:4647", # bj-semaphore
|
||||
"100.81.26.3:4647", # ash1d
|
||||
"100.103.147.94:4647", # ash2e
|
||||
"100.90.159.68:4647", # ch2
|
||||
"100.86.141.112:4647", # ch3
|
||||
"100.98.209.50:4647", # bj-onecloud1
|
||||
"100.120.225.29:4647" # de
|
||||
]
|
||||
host_volume "fnsync" {
|
||||
path = "/mnt/fnsync"
|
||||
read_only = false
|
||||
}
|
||||
# 禁用Docker驱动，只使用Podman
|
||||
options {
|
||||
"driver.raw_exec.enable" = "1"
|
||||
"driver.exec.enable" = "1"
|
||||
}
|
||||
}
|
||||
|
||||
# 配置Podman插件目录
|
||||
plugin_dir = "/opt/nomad/plugins"
|
||||
|
||||
addresses {
|
||||
http = "{{ ansible_host }}"
|
||||
rpc = "{{ ansible_host }}"
|
||||
serf = "{{ ansible_host }}"
|
||||
}
|
||||
|
||||
advertise {
|
||||
http = "{{ ansible_host }}:4646"
|
||||
rpc = "{{ ansible_host }}:4647"
|
||||
serf = "{{ ansible_host }}:4648"
|
||||
}
|
||||
|
||||
consul {
|
||||
address = "100.116.158.95:8500"
|
||||
}
|
||||
|
||||
# 配置Podman驱动
|
||||
plugin "podman" {
|
||||
config {
|
||||
volumes {
|
||||
enabled = true
|
||||
}
|
||||
logging {
|
||||
type = "journald"
|
||||
}
|
||||
gc {
|
||||
container = true
|
||||
}
|
||||
}
|
||||
}
|
||||
dest: "{{ nomad_config_dir }}/nomad.hcl"
|
||||
owner: root
|
||||
group: root
|
||||
mode: '0644'
|
||||
|
||||
- name: 启动Nomad服务
|
||||
systemd:
|
||||
name: nomad
|
||||
state: restarted
|
||||
enabled: yes
|
||||
daemon_reload: yes
|
||||
|
||||
- name: 检查Nomad服务状态
|
||||
command: systemctl status nomad
|
||||
register: nomad_status
|
||||
changed_when: false
|
||||
|
||||
- name: 显示Nomad服务状态
|
||||
debug:
|
||||
var: nomad_status.stdout_lines
|
||||
44
deployment/ansible/playbooks/configure-nomad-unified.yml
Normal file
44
deployment/ansible/playbooks/configure-nomad-unified.yml
Normal file
@@ -0,0 +1,44 @@
|
||||
---
|
||||
- name: 统一配置所有Nomad节点
|
||||
hosts: nomad_nodes
|
||||
become: yes
|
||||
|
||||
tasks:
|
||||
- name: 备份当前Nomad配置
|
||||
copy:
|
||||
src: /etc/nomad.d/nomad.hcl
|
||||
dest: /etc/nomad.d/nomad.hcl.bak
|
||||
remote_src: yes
|
||||
ignore_errors: yes
|
||||
|
||||
- name: 生成统一Nomad配置
|
||||
template:
|
||||
src: ../templates/nomad-unified.hcl.j2
|
||||
dest: /etc/nomad.d/nomad.hcl
|
||||
owner: root
|
||||
group: root
|
||||
mode: '0644'
|
||||
|
||||
- name: 重启Nomad服务
|
||||
systemd:
|
||||
name: nomad
|
||||
state: restarted
|
||||
enabled: yes
|
||||
daemon_reload: yes
|
||||
|
||||
- name: 等待Nomad服务就绪
|
||||
wait_for:
|
||||
port: 4646
|
||||
host: "{{ inventory_hostname }}.tailnet-68f9.ts.net"
|
||||
delay: 10
|
||||
timeout: 60
|
||||
ignore_errors: yes
|
||||
|
||||
- name: 检查Nomad服务状态
|
||||
command: systemctl status nomad
|
||||
register: nomad_status
|
||||
changed_when: false
|
||||
|
||||
- name: 显示Nomad服务状态
|
||||
debug:
|
||||
var: nomad_status.stdout_lines
|
||||
@@ -1,105 +0,0 @@
|
||||
---
|
||||
- name: 部署韩国节点Nomad配置
|
||||
hosts: ch2,ch3
|
||||
become: yes
|
||||
gather_facts: no
|
||||
vars:
|
||||
nomad_config_dir: "/etc/nomad.d"
|
||||
nomad_config_file: "{{ nomad_config_dir }}/nomad.hcl"
|
||||
source_config_dir: "/root/mgmt/infrastructure/configs/server"
|
||||
|
||||
tasks:
|
||||
- name: 获取主机名短名称（去掉.global后缀）
|
||||
set_fact:
|
||||
short_hostname: "{{ inventory_hostname | regex_replace('\\.global$', '') }}"
|
||||
|
||||
- name: 确保 Nomad 配置目录存在
|
||||
file:
|
||||
path: "{{ nomad_config_dir }}"
|
||||
state: directory
|
||||
owner: root
|
||||
group: root
|
||||
mode: '0755'
|
||||
|
||||
- name: 部署 Nomad 配置文件到韩国节点
|
||||
copy:
|
||||
src: "{{ source_config_dir }}/nomad-{{ short_hostname }}.hcl"
|
||||
dest: "{{ nomad_config_file }}"
|
||||
owner: root
|
||||
group: root
|
||||
mode: '0644'
|
||||
backup: yes
|
||||
notify: restart nomad
|
||||
|
||||
- name: 检查 Nomad 二进制文件位置
|
||||
shell: which nomad || find /usr -name nomad 2>/dev/null | head -1
|
||||
register: nomad_binary_path
|
||||
failed_when: nomad_binary_path.stdout == ""
|
||||
|
||||
- name: 创建/更新 Nomad systemd 服务文件
|
||||
copy:
|
||||
dest: "/etc/systemd/system/nomad.service"
|
||||
owner: root
|
||||
group: root
|
||||
mode: '0644'
|
||||
content: |
|
||||
[Unit]
|
||||
Description=Nomad
|
||||
Documentation=https://www.nomadproject.io/
|
||||
Requires=network-online.target
|
||||
After=network-online.target
|
||||
|
||||
[Service]
|
||||
Type=notify
|
||||
User=root
|
||||
Group=root
|
||||
ExecStart={{ nomad_binary_path.stdout }} agent -config=/etc/nomad.d/nomad.hcl
|
||||
ExecReload=/bin/kill -HUP $MAINPID
|
||||
KillMode=process
|
||||
Restart=on-failure
|
||||
LimitNOFILE=65536
|
||||
|
||||
[Install]
|
||||
WantedBy=multi-user.target
|
||||
notify: restart nomad
|
||||
|
||||
- name: 确保 Nomad 数据目录存在
|
||||
file:
|
||||
path: "/opt/nomad/data"
|
||||
state: directory
|
||||
owner: root
|
||||
group: root
|
||||
mode: '0755'
|
||||
|
||||
- name: 重新加载 systemd daemon
|
||||
systemd:
|
||||
daemon_reload: yes
|
||||
|
||||
- name: 启用并启动 Nomad 服务
|
||||
systemd:
|
||||
name: nomad
|
||||
enabled: yes
|
||||
state: started
|
||||
|
||||
- name: 等待 Nomad 服务启动
|
||||
wait_for:
|
||||
port: 4646
|
||||
host: "{{ ansible_host }}"
|
||||
delay: 5
|
||||
timeout: 30
|
||||
ignore_errors: yes
|
||||
|
||||
- name: 显示 Nomad 服务状态
|
||||
command: systemctl status nomad
|
||||
register: nomad_status
|
||||
changed_when: false
|
||||
|
||||
- name: 显示 Nomad 服务状态信息
|
||||
debug:
|
||||
var: nomad_status.stdout_lines
|
||||
|
||||
handlers:
|
||||
- name: restart nomad
|
||||
systemd:
|
||||
name: nomad
|
||||
state: restarted
|
||||
@@ -1,105 +0,0 @@
|
||||
---
|
||||
- name: 部署韩国节点Nomad配置
|
||||
hosts: ch2,ch3
|
||||
become: yes
|
||||
gather_facts: no
|
||||
vars:
|
||||
nomad_config_dir: "/etc/nomad.d"
|
||||
nomad_config_file: "{{ nomad_config_dir }}/nomad.hcl"
|
||||
source_config_dir: "/root/mgmt/infrastructure/configs/server"
|
||||
|
||||
tasks:
|
||||
- name: 获取主机名短名称（去掉后缀）
|
||||
set_fact:
|
||||
short_hostname: "{{ inventory_hostname | regex_replace('\\$', '') }}"
|
||||
|
||||
- name: 确保 Nomad 配置目录存在
|
||||
file:
|
||||
path: "{{ nomad_config_dir }}"
|
||||
state: directory
|
||||
owner: root
|
||||
group: root
|
||||
mode: '0755'
|
||||
|
||||
- name: 部署 Nomad 配置文件到韩国节点
|
||||
copy:
|
||||
src: "{{ source_config_dir }}/nomad-{{ short_hostname }}.hcl"
|
||||
dest: "{{ nomad_config_file }}"
|
||||
owner: root
|
||||
group: root
|
||||
mode: '0644'
|
||||
backup: yes
|
||||
notify: restart nomad
|
||||
|
||||
- name: 检查 Nomad 二进制文件位置
|
||||
shell: which nomad || find /usr -name nomad 2>/dev/null | head -1
|
||||
register: nomad_binary_path
|
||||
failed_when: nomad_binary_path.stdout == ""
|
||||
|
||||
- name: 创建/更新 Nomad systemd 服务文件
|
||||
copy:
|
||||
dest: "/etc/systemd/system/nomad.service"
|
||||
owner: root
|
||||
group: root
|
||||
mode: '0644'
|
||||
content: |
|
||||
[Unit]
|
||||
Description=Nomad
|
||||
Documentation=https://www.nomadproject.io/
|
||||
Requires=network-online.target
|
||||
After=network-online.target
|
||||
|
||||
[Service]
|
||||
Type=notify
|
||||
User=root
|
||||
Group=root
|
||||
ExecStart={{ nomad_binary_path.stdout }} agent -config=/etc/nomad.d/nomad.hcl
|
||||
ExecReload=/bin/kill -HUP $MAINPID
|
||||
KillMode=process
|
||||
Restart=on-failure
|
||||
LimitNOFILE=65536
|
||||
|
||||
[Install]
|
||||
WantedBy=multi-user.target
|
||||
notify: restart nomad
|
||||
|
||||
- name: 确保 Nomad 数据目录存在
|
||||
file:
|
||||
path: "/opt/nomad/data"
|
||||
state: directory
|
||||
owner: root
|
||||
group: root
|
||||
mode: '0755'
|
||||
|
||||
- name: 重新加载 systemd daemon
|
||||
systemd:
|
||||
daemon_reload: yes
|
||||
|
||||
- name: 启用并启动 Nomad 服务
|
||||
systemd:
|
||||
name: nomad
|
||||
enabled: yes
|
||||
state: started
|
||||
|
||||
- name: 等待 Nomad 服务启动
|
||||
wait_for:
|
||||
port: 4646
|
||||
host: "{{ ansible_host }}"
|
||||
delay: 5
|
||||
timeout: 30
|
||||
ignore_errors: yes
|
||||
|
||||
- name: 显示 Nomad 服务状态
|
||||
command: systemctl status nomad
|
||||
register: nomad_status
|
||||
changed_when: false
|
||||
|
||||
- name: 显示 Nomad 服务状态信息
|
||||
debug:
|
||||
var: nomad_status.stdout_lines
|
||||
|
||||
handlers:
|
||||
- name: restart nomad
|
||||
systemd:
|
||||
name: nomad
|
||||
state: restarted
|
||||
73
deployment/ansible/playbooks/fix-nomad-consul-roles.yml
Normal file
73
deployment/ansible/playbooks/fix-nomad-consul-roles.yml
Normal file
@@ -0,0 +1,73 @@
|
||||
---
|
||||
- name: 修正Nomad节点的Consul角色配置
|
||||
hosts: nomad_nodes
|
||||
become: yes
|
||||
vars:
|
||||
consul_addresses: "master.tailnet-68f9.ts.net:8500,ash3c.tailnet-68f9.ts.net:8500,warden.tailnet-68f9.ts.net:8500"
|
||||
|
||||
tasks:
|
||||
- name: 备份原始Nomad配置
|
||||
copy:
|
||||
src: /etc/nomad.d/nomad.hcl
|
||||
dest: /etc/nomad.d/nomad.hcl.bak_{{ ansible_date_time.iso8601 }}
|
||||
remote_src: yes
|
||||
|
||||
- name: 检查节点角色
|
||||
shell: grep -A 1 "server {" /etc/nomad.d/nomad.hcl | grep "enabled = true" | wc -l
|
||||
register: is_server
|
||||
changed_when: false
|
||||
|
||||
- name: 检查节点角色
|
||||
shell: grep -A 1 "client {" /etc/nomad.d/nomad.hcl | grep "enabled = true" | wc -l
|
||||
register: is_client
|
||||
changed_when: false
|
||||
|
||||
- name: 修正服务器节点的Consul配置
|
||||
blockinfile:
|
||||
path: /etc/nomad.d/nomad.hcl
|
||||
marker: "# {mark} ANSIBLE MANAGED BLOCK - CONSUL CONFIG"
|
||||
block: |
|
||||
consul {
|
||||
address = "{{ consul_addresses }}"
|
||||
server_service_name = "nomad"
|
||||
client_service_name = "nomad-client"
|
||||
auto_advertise = true
|
||||
server_auto_join = true
|
||||
client_auto_join = false
|
||||
}
|
||||
replace: true
|
||||
when: is_server.stdout == "1"
|
||||
|
||||
- name: 修正客户端节点的Consul配置
|
||||
blockinfile:
|
||||
path: /etc/nomad.d/nomad.hcl
|
||||
marker: "# {mark} ANSIBLE MANAGED BLOCK - CONSUL CONFIG"
|
||||
block: |
|
||||
consul {
|
||||
address = "{{ consul_addresses }}"
|
||||
server_service_name = "nomad"
|
||||
client_service_name = "nomad-client"
|
||||
auto_advertise = true
|
||||
server_auto_join = false
|
||||
client_auto_join = true
|
||||
}
|
||||
replace: true
|
||||
when: is_client.stdout == "1"
|
||||
|
||||
- name: 重启Nomad服务
|
||||
systemd:
|
||||
name: nomad
|
||||
state: restarted
|
||||
enabled: yes
|
||||
daemon_reload: yes
|
||||
|
||||
- name: 等待Nomad服务启动
|
||||
wait_for:
|
||||
port: 4646
|
||||
host: "{{ ansible_host }}"
|
||||
timeout: 30
|
||||
|
||||
- name: 显示节点角色和配置
|
||||
debug:
|
||||
msg: "节点 {{ inventory_hostname }} 是 {{ '服务器' if is_server.stdout == '1' else '客户端' }} 节点，Consul配置已更新"
|
||||
|
||||
43
deployment/ansible/playbooks/update-nomad-consul-config.yml
Normal file
43
deployment/ansible/playbooks/update-nomad-consul-config.yml
Normal file
@@ -0,0 +1,43 @@
|
||||
---
|
||||
- name: 更新所有Nomad节点的Consul配置
|
||||
hosts: nomad_nodes
|
||||
become: yes
|
||||
vars:
|
||||
consul_addresses: "master.tailnet-68f9.ts.net:8500,ash3c.tailnet-68f9.ts.net:8500,warden.tailnet-68f9.ts.net:8500"
|
||||
|
||||
tasks:
|
||||
- name: 备份原始Nomad配置
|
||||
copy:
|
||||
src: /etc/nomad.d/nomad.hcl
|
||||
dest: /etc/nomad.d/nomad.hcl.backup.{{ ansible_date_time.epoch }}
|
||||
remote_src: yes
|
||||
backup: yes
|
||||
|
||||
- name: 更新Nomad Consul配置
|
||||
lineinfile:
|
||||
path: /etc/nomad.d/nomad.hcl
|
||||
regexp: '^\s*address\s*=\s*".*"'
|
||||
line: ' address = "{{ consul_addresses }}"'
|
||||
state: present
|
||||
|
||||
- name: 重启Nomad服务
|
||||
systemd:
|
||||
name: nomad
|
||||
state: restarted
|
||||
enabled: yes
|
||||
daemon_reload: 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: "节点 {{ inventory_hostname }} Nomad服务状态: {{ nomad_status.status.ActiveState }}"
|
||||
Reference in New Issue
Block a user