Service monitor

Service Monitor

Overview

Check the health of services, APIs, and infrastructure components. Verify uptime, response times, and correct behavior of endpoints. Produce a service health dashboard.

Process

Step 1: Identify Services to Monitor

Ask the user OR auto-detect:

Auto-detect running services:

Platform	Command
Windows	Get-Service | Where-Object {$_.Status -eq 'Running'} | Select Name, DisplayName
Linux	systemctl list-units --type=service --state=running --no-pager
macOS	launchctl list | grep -v "^-|com.apple"

Auto-detect listening ports:

Platform	Command
Windows	Get-NetTCPConnection -State Listen | Select LocalPort, @{N='Process';E={(Get-Process -Id $_.OwningProcess -ErrorAction SilentlyContinue).Name}} | Sort LocalPort -Unique
Linux	ss -tlnp | tail -n +2
macOS	lsof -iTCP -sTCP:LISTEN -n -P | awk '{print $1, $9}'

Step 2: Health Checks

HTTP/HTTPS endpoints:

# Basic availability + response time
curl -s -o /dev/null -w "HTTP %{http_code} | Time: %{time_total}s | Size: %{size_download}B" [url]

# With timeout
curl -s --max-time 10 -o /dev/null -w "%{http_code}" [url]

# Follow redirects
curl -sL -o /dev/null -w "%{http_code} %{url_effective}" [url]

# Check specific header
curl -sI [url] | grep -i [header]

TCP port checks:

Platform	Command
Windows	Test-NetConnection -ComputerName [host] -Port [port] -InformationLevel Quiet
Linux	timeout 5 bash -c "echo > /dev/tcp/[host]/[port]" 2>/dev/null && echo "OPEN" || echo "CLOSED"
macOS	nc -z -w5 [host] [port] 2>&1

Database connectivity:

Database	Check Command
PostgreSQL	pg_isready -h [host] -p [port]
MySQL	mysqladmin ping -h [host] -P [port] 2>&1
Redis	redis-cli -h [host] -p [port] ping
MongoDB	mongosh --host [host] --port [port] --eval "db.runCommand({ping:1})" --quiet

Service-specific checks:

Service	Check
DNS	dig @[server] [domain] +short +time=5
SMTP	nc -z -w5 [host] 25 || nc -z -w5 [host] 587
NTP	ntpdate -q [server] 2>&1 || timedatectl show-timesync
LDAP	ldapsearch -x -H ldap://[host] -b "" -s base 2>&1 | head -5

Step 3: Response Time Analysis

Run multiple checks to assess consistency:

# 5 requests to measure consistency
for i in 1 2 3 4 5; do
  curl -s -o /dev/null -w "%{time_total}\n" [url]
  sleep 1
done

Response time thresholds:

Response Time	Rating	Action
< 200ms	Excellent	No action
200-500ms	Good	Monitor
500ms-1s	Slow	Investigate
1s-5s	Poor	Needs optimization
> 5s	Critical	Immediate attention
Timeout	Down	Alert/escalate

Step 4: Process Health

Check if process is running:

Platform	Command
Windows	Get-Process -Name [name] -ErrorAction SilentlyContinue | Select Name, Id, CPU, @{N='Mem(MB)';E={[math]::Round($_.WorkingSet64/1MB)}}
Linux	pgrep -a [name]
macOS	pgrep -l [name]

Process resource consumption:

Platform	Command
Windows	Get-Process -Name [name] | Select Id, CPU, WorkingSet64, HandleCount, StartTime
Linux	ps -p $(pgrep [name]) -o pid,pcpu,pmem,vsz,rss,etime

Check for zombie/defunct processes:

Platform	Command
Linux	ps aux | grep -w Z | grep -v grep

Step 5: Dependency Chain Check

Map and verify service dependencies:

Service A → depends on → Service B → depends on → Database
   ✓                         ✓                        ?

For each dependency:

1. Is it running?
2. Is it responding?
3. Is the response correct (not just 200 OK)?

Step 6: Present Health Dashboard

## Service Health Dashboard
**Checked:** [timestamp] | **Host:** [hostname]

### Service Status
| Service | Port | Status | Response | Latency |
|---------|------|--------|----------|---------|
| [name] | [port] | [UP/DOWN] | [code] | [ms] |
| [name] | [port] | [UP/DOWN] | [code] | [ms] |

### Summary
- Total services: [N]
- Healthy: [N] ✓
- Degraded: [N] ⚠
- Down: [N] ✗

### Issues Detected
1. [Service X] - [issue description]
2. [Service Y] - [issue description]

### Recommendations
- [Action items for degraded/down services]

Rules

• NEVER send credentials in health check commands
• NEVER modify service configuration
• Use timeouts on ALL network checks (max 10s)
• If a service requires auth to check, ask user for method
• Don't flood services with requests (max 5 checks per endpoint)
• Report response codes accurately (don't assume 200 = healthy)
• Check from the host's perspective, not externally