1 feat: 重构基础设施架构并完善Consul集群配置
2
3 主要变更:
4 - 重构Terraform/OpenTofu目录结构，统一迁移至infrastructure/opentofu
5 - 添加"7天创造世界"文档，记录基础设施建设演进逻辑
6 - 更新Consul集群配置管理经验，添加实际案例和解决方案
7 - 修正README中的Sticky Note，反映Consul集群健康状态
8 - 添加Ansible部署配置和inventory文件
9 - 完善项目文档结构，添加各组件配置指南
10
11 技术架构演进:
12 - 第1天: Tailscale网络连接基础 ✅
13 - 第2天: Ansible分布式控制 ✅
14 - 第3天: Nomad服务感知与任务调度 ✅
15 - 第4天: Consul配置集中管理 ✅
16 - 第5天: OpenTofu状态一致性 ✅
17 - 第6天: Vault密钥管理 ⏳
18 - 第7天: Waypoint应用部署 ⏳
This commit is contained in:
168
deployment/ansible/playbooks/disk/disk-analysis-ncdu.yml
Normal file
168
deployment/ansible/playbooks/disk/disk-analysis-ncdu.yml
Normal file
@@ -0,0 +1,168 @@
|
||||
---
|
||||
- name: 磁盘空间分析 - 使用 ncdu 工具
|
||||
hosts: all
|
||||
become: yes
|
||||
vars:
|
||||
ncdu_scan_paths:
|
||||
- "/"
|
||||
- "/var"
|
||||
- "/opt"
|
||||
- "/home"
|
||||
output_dir: "/tmp/disk-analysis"
|
||||
|
||||
tasks:
|
||||
- name: 安装 ncdu 工具
|
||||
package:
|
||||
name: ncdu
|
||||
state: present
|
||||
register: ncdu_install
|
||||
|
||||
- name: 创建输出目录
|
||||
file:
|
||||
path: "{{ output_dir }}"
|
||||
state: directory
|
||||
mode: '0755'
|
||||
|
||||
- name: 检查磁盘空间使用情况
|
||||
shell: df -h
|
||||
register: disk_usage
|
||||
|
||||
- name: 显示当前磁盘使用情况
|
||||
debug:
|
||||
msg: |
|
||||
=== {{ inventory_hostname }} 磁盘使用情况 ===
|
||||
{{ disk_usage.stdout }}
|
||||
|
||||
- name: 使用 ncdu 扫描根目录并生成报告
|
||||
shell: |
|
||||
ncdu -x -o {{ output_dir }}/ncdu-root-{{ inventory_hostname }}.json /
|
||||
async: 300
|
||||
poll: 0
|
||||
register: ncdu_root_scan
|
||||
|
||||
- name: 使用 ncdu 扫描 /var 目录
|
||||
shell: |
|
||||
ncdu -x -o {{ output_dir }}/ncdu-var-{{ inventory_hostname }}.json /var
|
||||
async: 180
|
||||
poll: 0
|
||||
register: ncdu_var_scan
|
||||
when: ansible_mounts | selectattr('mount', 'equalto', '/var') | list | length > 0 or '/var' in ansible_mounts | map(attribute='mount') | list
|
||||
|
||||
- name: 使用 ncdu 扫描 /opt 目录
|
||||
shell: |
|
||||
ncdu -x -o {{ output_dir }}/ncdu-opt-{{ inventory_hostname }}.json /opt
|
||||
async: 120
|
||||
poll: 0
|
||||
register: ncdu_opt_scan
|
||||
when: ansible_mounts | selectattr('mount', 'equalto', '/opt') | list | length > 0 or '/opt' in ansible_mounts | map(attribute='mount') | list
|
||||
|
||||
- name: 等待根目录扫描完成
|
||||
async_status:
|
||||
jid: "{{ ncdu_root_scan.ansible_job_id }}"
|
||||
register: ncdu_root_result
|
||||
until: ncdu_root_result.finished
|
||||
retries: 60
|
||||
delay: 5
|
||||
|
||||
- name: 等待 /var 目录扫描完成
|
||||
async_status:
|
||||
jid: "{{ ncdu_var_scan.ansible_job_id }}"
|
||||
register: ncdu_var_result
|
||||
until: ncdu_var_result.finished
|
||||
retries: 36
|
||||
delay: 5
|
||||
when: ncdu_var_scan is defined and ncdu_var_scan.ansible_job_id is defined
|
||||
|
||||
- name: 等待 /opt 目录扫描完成
|
||||
async_status:
|
||||
jid: "{{ ncdu_opt_scan.ansible_job_id }}"
|
||||
register: ncdu_opt_result
|
||||
until: ncdu_opt_result.finished
|
||||
retries: 24
|
||||
delay: 5
|
||||
when: ncdu_opt_scan is defined and ncdu_opt_scan.ansible_job_id is defined
|
||||
|
||||
- name: 生成磁盘使用分析报告
|
||||
shell: |
|
||||
echo "=== {{ inventory_hostname }} 磁盘分析报告 ===" > {{ output_dir }}/disk-report-{{ inventory_hostname }}.txt
|
||||
echo "生成时间: $(date)" >> {{ output_dir }}/disk-report-{{ inventory_hostname }}.txt
|
||||
echo "" >> {{ output_dir }}/disk-report-{{ inventory_hostname }}.txt
|
||||
echo "=== 磁盘使用情况 ===" >> {{ output_dir }}/disk-report-{{ inventory_hostname }}.txt
|
||||
df -h >> {{ output_dir }}/disk-report-{{ inventory_hostname }}.txt
|
||||
echo "" >> {{ output_dir }}/disk-report-{{ inventory_hostname }}.txt
|
||||
echo "=== 最大的目录 (前10个) ===" >> {{ output_dir }}/disk-report-{{ inventory_hostname }}.txt
|
||||
du -h --max-depth=2 / 2>/dev/null | sort -hr | head -10 >> {{ output_dir }}/disk-report-{{ inventory_hostname }}.txt
|
||||
echo "" >> {{ output_dir }}/disk-report-{{ inventory_hostname }}.txt
|
||||
echo "=== /var 目录最大文件 ===" >> {{ output_dir }}/disk-report-{{ inventory_hostname }}.txt
|
||||
find /var -type f -size +100M -exec ls -lh {} \; 2>/dev/null | head -10 >> {{ output_dir }}/disk-report-{{ inventory_hostname }}.txt
|
||||
echo "" >> {{ output_dir }}/disk-report-{{ inventory_hostname }}.txt
|
||||
echo "=== /tmp 目录使用情况 ===" >> {{ output_dir }}/disk-report-{{ inventory_hostname }}.txt
|
||||
du -sh /tmp/* 2>/dev/null | sort -hr | head -5 >> {{ output_dir }}/disk-report-{{ inventory_hostname }}.txt
|
||||
echo "" >> {{ output_dir }}/disk-report-{{ inventory_hostname }}.txt
|
||||
echo "=== 日志文件大小 ===" >> {{ output_dir }}/disk-report-{{ inventory_hostname }}.txt
|
||||
find /var/log -name "*.log" -type f -size +50M -exec ls -lh {} \; 2>/dev/null >> {{ output_dir }}/disk-report-{{ inventory_hostname }}.txt
|
||||
|
||||
- name: 显示分析报告
|
||||
shell: cat {{ output_dir }}/disk-report-{{ inventory_hostname }}.txt
|
||||
register: disk_report
|
||||
|
||||
- name: 输出磁盘分析结果
|
||||
debug:
|
||||
msg: "{{ disk_report.stdout }}"
|
||||
|
||||
- name: 检查是否有磁盘使用率超过 80%
|
||||
shell: df -h | awk 'NR>1 {gsub(/%/, "", $5); if($5 > 80) print $0}'
|
||||
register: high_usage_disks
|
||||
|
||||
- name: 警告高磁盘使用率
|
||||
debug:
|
||||
msg: |
|
||||
⚠️ 警告: {{ inventory_hostname }} 发现高磁盘使用率!
|
||||
{{ high_usage_disks.stdout }}
|
||||
when: high_usage_disks.stdout != ""
|
||||
|
||||
- name: 创建清理建议
|
||||
shell: |
|
||||
echo "=== {{ inventory_hostname }} 清理建议 ===" > {{ output_dir }}/cleanup-suggestions-{{ inventory_hostname }}.txt
|
||||
echo "" >> {{ output_dir }}/cleanup-suggestions-{{ inventory_hostname }}.txt
|
||||
echo "1. 检查日志文件:" >> {{ output_dir }}/cleanup-suggestions-{{ inventory_hostname }}.txt
|
||||
find /var/log -name "*.log" -type f -size +100M -exec echo " 大日志文件: {}" \; 2>/dev/null >> {{ output_dir }}/cleanup-suggestions-{{ inventory_hostname }}.txt
|
||||
echo "" >> {{ output_dir }}/cleanup-suggestions-{{ inventory_hostname }}.txt
|
||||
echo "2. 检查临时文件:" >> {{ output_dir }}/cleanup-suggestions-{{ inventory_hostname }}.txt
|
||||
find /tmp -type f -size +50M -exec echo " 大临时文件: {}" \; 2>/dev/null >> {{ output_dir }}/cleanup-suggestions-{{ inventory_hostname }}.txt
|
||||
echo "" >> {{ output_dir }}/cleanup-suggestions-{{ inventory_hostname }}.txt
|
||||
echo "3. 检查包缓存:" >> {{ output_dir }}/cleanup-suggestions-{{ inventory_hostname }}.txt
|
||||
if [ -d /var/cache/apt ]; then
|
||||
echo " APT 缓存大小: $(du -sh /var/cache/apt 2>/dev/null | cut -f1)" >> {{ output_dir }}/cleanup-suggestions-{{ inventory_hostname }}.txt
|
||||
fi
|
||||
if [ -d /var/cache/yum ]; then
|
||||
echo " YUM 缓存大小: $(du -sh /var/cache/yum 2>/dev/null | cut -f1)" >> {{ output_dir }}/cleanup-suggestions-{{ inventory_hostname }}.txt
|
||||
fi
|
||||
echo "" >> {{ output_dir }}/cleanup-suggestions-{{ inventory_hostname }}.txt
|
||||
echo "4. 检查容器相关:" >> {{ output_dir }}/cleanup-suggestions-{{ inventory_hostname }}.txt
|
||||
if command -v podman >/dev/null 2>&1; then
|
||||
echo " Podman 镜像: $(podman images --format 'table {{.Repository}} {{.Tag}} {{.Size}}' 2>/dev/null | wc -l) 个" >> {{ output_dir }}/cleanup-suggestions-{{ inventory_hostname }}.txt
|
||||
echo " Podman 容器: $(podman ps -a --format 'table {{.Names}} {{.Status}}' 2>/dev/null | wc -l) 个" >> {{ output_dir }}/cleanup-suggestions-{{ inventory_hostname }}.txt
|
||||
fi
|
||||
|
||||
- name: 显示清理建议
|
||||
shell: cat {{ output_dir }}/cleanup-suggestions-{{ inventory_hostname }}.txt
|
||||
register: cleanup_suggestions
|
||||
|
||||
- name: 输出清理建议
|
||||
debug:
|
||||
msg: "{{ cleanup_suggestions.stdout }}"
|
||||
|
||||
- name: 保存 ncdu 文件位置信息
|
||||
debug:
|
||||
msg: |
|
||||
📁 ncdu 扫描文件已保存到:
|
||||
- 根目录: {{ output_dir }}/ncdu-root-{{ inventory_hostname }}.json
|
||||
- /var 目录: {{ output_dir }}/ncdu-var-{{ inventory_hostname }}.json (如果存在)
|
||||
- /opt 目录: {{ output_dir }}/ncdu-opt-{{ inventory_hostname }}.json (如果存在)
|
||||
|
||||
💡 使用方法:
|
||||
ncdu -f {{ output_dir }}/ncdu-root-{{ inventory_hostname }}.json
|
||||
|
||||
📊 完整报告: {{ output_dir }}/disk-report-{{ inventory_hostname }}.txt
|
||||
🧹 清理建议: {{ output_dir }}/cleanup-suggestions-{{ inventory_hostname }}.txt
|
||||
96
deployment/ansible/playbooks/disk/disk-cleanup.yml
Normal file
96
deployment/ansible/playbooks/disk/disk-cleanup.yml
Normal file
@@ -0,0 +1,96 @@
|
||||
---
|
||||
- name: 磁盘清理工具
|
||||
hosts: all
|
||||
become: yes
|
||||
vars:
|
||||
cleanup_logs: true
|
||||
cleanup_cache: true
|
||||
cleanup_temp: true
|
||||
cleanup_containers: false # 谨慎操作
|
||||
|
||||
tasks:
|
||||
- name: 检查磁盘使用情况 (清理前)
|
||||
shell: df -h
|
||||
register: disk_before
|
||||
|
||||
- name: 显示清理前磁盘使用情况
|
||||
debug:
|
||||
msg: |
|
||||
=== {{ inventory_hostname }} 清理前磁盘使用情况 ===
|
||||
{{ disk_before.stdout }}
|
||||
|
||||
- name: 清理系统日志 (保留最近7天)
|
||||
shell: |
|
||||
journalctl --vacuum-time=7d
|
||||
find /var/log -name "*.log" -type f -mtime +7 -exec truncate -s 0 {} \;
|
||||
find /var/log -name "*.log.*" -type f -mtime +7 -delete
|
||||
when: cleanup_logs | bool
|
||||
register: log_cleanup
|
||||
|
||||
- name: 清理包管理器缓存
|
||||
block:
|
||||
- name: 清理 APT 缓存 (Debian/Ubuntu)
|
||||
shell: |
|
||||
apt-get clean
|
||||
apt-get autoclean
|
||||
apt-get autoremove -y
|
||||
when: ansible_os_family == "Debian"
|
||||
|
||||
- name: 清理 YUM/DNF 缓存 (RedHat/CentOS)
|
||||
shell: |
|
||||
if command -v dnf >/dev/null 2>&1; then
|
||||
dnf clean all
|
||||
elif command -v yum >/dev/null 2>&1; then
|
||||
yum clean all
|
||||
fi
|
||||
when: ansible_os_family == "RedHat"
|
||||
when: cleanup_cache | bool
|
||||
|
||||
- name: 清理临时文件
|
||||
shell: |
|
||||
find /tmp -type f -atime +7 -delete 2>/dev/null || true
|
||||
find /var/tmp -type f -atime +7 -delete 2>/dev/null || true
|
||||
rm -rf /tmp/.* 2>/dev/null || true
|
||||
when: cleanup_temp | bool
|
||||
|
||||
- name: 清理 Podman 资源 (谨慎操作)
|
||||
block:
|
||||
- name: 停止所有容器
|
||||
shell: podman stop --all
|
||||
ignore_errors: yes
|
||||
|
||||
- name: 删除未使用的容器
|
||||
shell: podman container prune -f
|
||||
ignore_errors: yes
|
||||
|
||||
- name: 删除未使用的镜像
|
||||
shell: podman image prune -f
|
||||
ignore_errors: yes
|
||||
|
||||
- name: 删除未使用的卷
|
||||
shell: podman volume prune -f
|
||||
ignore_errors: yes
|
||||
when: cleanup_containers | bool
|
||||
|
||||
- name: 清理核心转储文件
|
||||
shell: |
|
||||
find /var/crash -name "core.*" -type f -delete 2>/dev/null || true
|
||||
find / -name "core" -type f -size +10M -delete 2>/dev/null || true
|
||||
ignore_errors: yes
|
||||
|
||||
- name: 检查磁盘使用情况 (清理后)
|
||||
shell: df -h
|
||||
register: disk_after
|
||||
|
||||
- name: 显示清理结果
|
||||
debug:
|
||||
msg: |
|
||||
=== {{ inventory_hostname }} 清理完成 ===
|
||||
|
||||
清理前:
|
||||
{{ disk_before.stdout }}
|
||||
|
||||
清理后:
|
||||
{{ disk_after.stdout }}
|
||||
|
||||
🧹 清理操作完成!
|
||||
Reference in New Issue
Block a user