Conduct periodic access reviews and certifications. Implement access governance and recertification workflows. Use when managing access compliance.
日本語の概要は準備中です。原文の説明を表示しています。
Create and manage systemd services and timers. Configure service dependencies and resource limits. Use when managing system services.
インストールする前に、エージェントに与えられる指示の中身を確認できます。
Create, manage, and monitor systemd services and timers. Covers unit file authoring, dependency management, socket activation, resource limits, journalctl log analysis, and production hardening.
# /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
# 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
# /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
# /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
# 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
# /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
# 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
# /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
# 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
| Symptom | Diagnostic Command | Common Fix |
|---|---|---|
| Service fails to start | systemctl status myapp, journalctl -u myapp -n 50 | Check ExecStart path, permissions, config syntax |
| Service keeps restarting | journalctl -u myapp --since "5 min ago" | Check StartLimitBurst; look for crash in logs |
| "Main process exited, code=exited, status=217" | journalctl -u myapp | User or group in unit file does not exist |
| "Failed to set up mount namespacing" | Check ProtectSystem/PrivateTmp | Kernel too old or SELinux blocking; relax directives |
| Timer not firing | systemctl list-timers, systemctl status backup.timer | Ensure timer is enabled; validate OnCalendar expression |
| Service starts before dependency | Check After= and Requires= | Add After=dependency.service for ordering |
| OOM killed | journalctl -k | grep oom, dmesg | Increase MemoryMax or optimize application memory |
| Cannot bind to port 80 | Check AmbientCapabilities | Add CAP_NET_BIND_SERVICE or use a higher port |
linux-administration -- General system administration contextperformance-tuning -- Kernel tuning and resource optimizationuser-management -- Service accounts and permissionsbackup-recovery -- Scheduling backups with systemd timersまだレビューはありません。使ってみた感想をお寄せください。
概要と使いどころ
Conduct periodic access reviews and certifications. Implement access governance and recertification workflows. Use when managing access compliance.
日本語の概要は準備中です。原文の説明を表示しています。
Build automated evaluation suites for AI agents using golden datasets, rubrics, and regression gates. Use when shipping agent features, validating prompt changes, or gating deployments on quality.
日本語の概要は準備中です。原文の説明を表示しています。
Instrument AI agents with tracing, token metrics, latency, and cost visibility. Use for reliability and debugging.
日本語の概要は準備中です。原文の説明を表示しています。
Secure AI agents against prompt injection, tool abuse, and data exfiltration with defense-in-depth controls. Use when building, deploying, or hardening agentic AI systems that invoke tools, access data, or interact with production infrastructure.
日本語の概要は準備中です。原文の説明を表示しています。
Secure AI coding agents (Claude Code, Cursor, Codex, Copilot) with permission boundaries, secret protection, code review gates, and safe sandbox configurations for team environments.
日本語の概要は準備中です。原文の説明を表示しています。
Use service mesh patterns for AI inference traffic management, mTLS, canary releases, policy enforcement, and cross-cluster resilience.
日本語の概要は準備中です。原文の説明を表示しています。