systemd-services

Systemd 服务

创建、管理和监控 systemd 服务与定时器。涵盖单元文件编写、依赖管理、套接字激活、资源限制、journalctl 日志分析和生产环境加固。

适用场景

• 将应用部署为受管理的后台服务
• 用 systemd 定时器替代 cron 作业，以便更好地记录日志和控制依赖关系
• 设置套接字激活，实现按需启动服务
• 为服务配置资源限制（CPU、内存、I/O）
• 调试服务启动失败和运行时崩溃
• 管理服务依赖关系和启动顺序

前置条件

• 运行 systemd 的 Linux 系统（大多数现代发行版均符合）
• 创建系统级单元文件所需的 root 或 sudo 权限
• 要作为服务运行的应用二进制文件或脚本
• 了解应用的启动/停止生命周期

服务单元文件——完整示例

# /etc/systemd/system/myapp.service
[Unit]
Description=MyApp Production Server
Documentation=https://docs.example.com/myapp
After=network-online.target postgresql.service
Wants=network-online.target
Requires=postgresql.service

[Service]
Type=notify
User=myapp
Group=myapp
WorkingDirectory=/opt/myapp

# Environment configuration
EnvironmentFile=/etc/myapp/env
Environment=NODE_ENV=production
Environment=PORT=8080

# Execution
ExecStartPre=/opt/myapp/bin/migrate --check
ExecStart=/opt/myapp/bin/server --config /etc/myapp/config.yaml
ExecStartPost=/opt/myapp/bin/healthcheck.sh
ExecReload=/bin/kill -HUP $MAINPID
ExecStop=/opt/myapp/bin/graceful-stop.sh

# Restart behavior
Restart=on-failure
RestartSec=5
StartLimitIntervalSec=300
StartLimitBurst=5

# Timeouts
TimeoutStartSec=30
TimeoutStopSec=30
WatchdogSec=60

# Security hardening
NoNewPrivileges=true
ProtectSystem=strict
ProtectHome=true
PrivateTmp=true
ReadWritePaths=/var/lib/myapp /var/log/myapp
CapabilityBoundingSet=
AmbientCapabilities=

# Logging
StandardOutput=journal
StandardError=journal
SyslogIdentifier=myapp

[Install]
WantedBy=multi-user.target

服务管理命令

# Reload systemd after creating or modifying unit files
systemctl daemon-reload

# Start, stop, restart a service
systemctl start myapp
systemctl stop myapp
systemctl restart myapp

# Reload service configuration without restart (if supported)
systemctl reload myapp

# Enable service to start on boot
systemctl enable myapp

# Enable and start in one command
systemctl enable --now myapp

# Disable and stop
systemctl disable --now myapp

# Check service status
systemctl status myapp

# Check if a service is active, enabled, or failed
systemctl is-active myapp
systemctl is-enabled myapp
systemctl is-failed myapp

# List all running services
systemctl list-units --type=service --state=running

# List all failed services
systemctl list-units --type=service --state=failed

# Show all properties of a service
systemctl show myapp

# Show specific property values
systemctl show myapp -p MainPID,MemoryCurrent,CPUUsageNSec

# Mask a service (prevent it from being started at all)
systemctl mask myapp

# Unmask
systemctl unmask myapp

# Reset a failed service state
systemctl reset-failed myapp

定时器单元（替代 Cron）

定时器文件

# /etc/systemd/system/backup.timer
[Unit]
Description=Daily backup timer

[Timer]
# Run daily at 2:30 AM
OnCalendar=*-*-* 02:30:00
# If the system was off at the scheduled time, run when it boots
Persistent=true
# Add random delay up to 15 minutes to avoid thundering herd
RandomizedDelaySec=900
# Associate with a specific service (defaults to same name .service)
Unit=backup.service

[Install]
WantedBy=timers.target

对应的服务文件

# /etc/systemd/system/backup.service
[Unit]
Description=Daily backup job
After=network-online.target
Wants=network-online.target

[Service]
Type=oneshot
User=backup
ExecStart=/usr/local/bin/run-backup.sh
StandardOutput=journal
StandardError=journal

定时器管理

# Common OnCalendar expressions:
# minutely, hourly, daily, weekly, monthly
# *-*-* 06:00:00       Daily at 6 AM
# Mon..Fri *-*-* 09:00 Weekdays at 9 AM
# *:0/15               Every 15 minutes

# Validate calendar expressions
systemd-analyze calendar "Mon..Fri *-*-* 09:00"

# List all active timers
systemctl list-timers --all

# Enable and start a timer
systemctl enable --now backup.timer

# Run the associated service immediately (for testing)
systemctl start backup.service

套接字激活

# /etc/systemd/system/myapp.socket
[Unit]
Description=MyApp Socket

[Socket]
ListenStream=8080
Accept=no
# Optionally bind to a specific IP
# ListenStream=10.0.1.10:8080

[Install]
WantedBy=sockets.target

# /etc/systemd/system/myapp.service
[Unit]
Description=MyApp Server
Requires=myapp.socket

[Service]
Type=notify
User=myapp
ExecStart=/opt/myapp/bin/server
# Service receives the socket file descriptor from systemd

[Install]
WantedBy=multi-user.target

# Enable the socket (service starts on first connection)
systemctl enable --now myapp.socket

# Check socket status
systemctl status myapp.socket

# List all listening sockets
systemctl list-sockets

依赖管理

# Key [Unit] directives for ordering and dependencies:
# After=            Start after these units (ordering only)
# Requires=         Hard dependency -- fail if this unit cannot start
# Wants=            Soft dependency -- try to start, don't fail if unavailable
# PartOf=           Stop this unit when the parent stops
# Conflicts=        Cannot run alongside this unit

# Visualize the dependency tree for a service
systemctl list-dependencies myapp

# Show reverse dependencies (who depends on this unit)
systemctl list-dependencies myapp --reverse

# Analyze boot order for a service
systemd-analyze critical-chain myapp.service

资源限制（cgroups v2）

# /etc/systemd/system/myapp.service.d/limits.conf
# (drop-in override file)
[Service]
# Memory limits
MemoryMax=1G
MemoryHigh=768M

# CPU limits
CPUQuota=200%          # Up to 2 full CPU cores
CPUWeight=100          # Relative weight (default=100)

# I/O limits
IOWeight=50
IOReadBandwidthMax=/dev/sda 100M
IOWriteBandwidthMax=/dev/sda 50M

# Process limits
LimitNOFILE=65535
LimitNPROC=4096
TasksMax=512

# Disable OOM killer (let the app handle it)
OOMPolicy=continue

# Apply drop-in overrides without editing the main unit file
mkdir -p /etc/systemd/system/myapp.service.d/

cat <<'EOF' > /etc/systemd/system/myapp.service.d/limits.conf
[Service]
MemoryMax=1G
CPUQuota=200%
EOF

systemctl daemon-reload
systemctl restart myapp

# View current resource usage for a service
systemctl status myapp                  # Shows Memory and CPU
systemd-cgtop                          # Real-time cgroup resource usage

# Edit a service's overrides interactively
systemctl edit myapp
# This creates a drop-in file automatically

Journalctl 日志分析

# Follow logs for a service in real time
journalctl -u myapp -f

# Show logs since last boot
journalctl -u myapp -b

# Show logs for a specific time range
journalctl -u myapp --since "2025-01-15 08:00" --until "2025-01-15 12:00"

# Show only error and above
journalctl -u myapp -p err

# Show the last 100 lines with full messages (no truncation)
journalctl -u myapp -n 100 --no-pager -l

# Show logs in JSON format (for parsing)
journalctl -u myapp -o json-pretty --no-pager | head -50

# Check journal disk usage and vacuum old entries
journalctl --disk-usage
journalctl --rotate
journalctl --vacuum-time=7d
journalctl --vacuum-size=500M

故障排查

症状	诊断命令	常见修复方法
服务无法启动	systemctl status myapp、journalctl -u myapp -n 50	检查 ExecStart 路径、权限和配置语法
服务不断重启	journalctl -u myapp --since "5 min ago"	检查 StartLimitBurst；在日志中查找崩溃信息
"Main process exited, code=exited, status=217"	journalctl -u myapp	单元文件中的用户或组不存在
"Failed to set up mount namespacing"	检查 ProtectSystem/PrivateTmp	内核版本过旧或 SELinux 阻止了操作；放宽相关指令
定时器未触发	systemctl list-timers、systemctl status backup.timer	确保定时器已启用；验证 OnCalendar 表达式
服务在依赖项之前启动	检查 After= 和 Requires=	添加 After=dependency.service 来指定顺序
被 OOM 杀死	journalctl -k | grep oom、dmesg	增大 MemoryMax 或优化应用内存使用
无法绑定到端口 80	检查 AmbientCapabilities	添加 CAP_NET_BIND_SERVICE 或使用更高的端口

相关技能

• linux-administration -- 通用系统管理背景知识
• performance-tuning -- 内核调优和资源优化
• user-management -- 服务账户和权限
• backup-recovery -- 使用 systemd 定时器安排备份