VIBE Monitor

VIBE module monitoring integration, including Claude Code observability and Dashboard usage.

VIBE module provides complete observability support, especially Claude Code’s OpenTelemetry integration for monitoring AI coding assistant usage via Grafana Dashboard.

Monitoring Architecture

Claude Code CLI
    │
    ├── OpenTelemetry Metrics ──► VictoriaMetrics (8428)
    └── OpenTelemetry Logs    ──► VictoriaLogs    (9428)
                                       │
                                       ▼
                               Grafana Dashboard
                               (Claude Code Dashboard)

Claude Code Observability

OpenTelemetry Config

Claude Code includes OpenTelemetry by default, auto-pushing metrics and logs to local VictoriaMetrics/VictoriaLogs:

Config file ~/.claude/settings.json:

{
  "otelEndpointMetrics": "http://127.0.0.1:8428/opentelemetry/v1/metrics",
  "otelEndpointLogs": "http://127.0.0.1:9428/insert/opentelemetry/v1/logs",
  "otelResourceAttributes": {
    "ip": "10.10.10.10",
    "job": "claude"
  }
}

Monitoring Metrics

Claude Code reports these main metrics:

Metric TypeDescription
SessionSession count, duration
TokenInput/output token count, cost stats
API CallsRequest count, latency, error rate
Tool UsageTool call counts (Read, Write, Bash, etc.)

Log Data

Claude Code logs include:

  • Session start/end events
  • User prompts and assistant responses
  • Tool call details
  • Errors and warnings

Grafana Dashboard

Access Dashboard

Claude Code Dashboard accessible at:

https://<ip>:3000/d/claude-code

Login with Grafana credentials (default: admin / pigsty).

Dashboard Features

PanelDescription
Session OverviewActive sessions, total sessions
Token UsageInput/output token trends, cost analysis
API PerformanceRequest latency, error rate
Tool Usage StatsTool call frequency, duration distribution
Session LogsDetailed session records and conversations

System Service Monitoring

Using systemd Monitoring

# Check Code-Server status
systemctl status code-server

# Check JupyterLab status
systemctl status jupyter

# View service logs
journalctl -u code-server -f
journalctl -u jupyter -f

Port Checks

# Check Code-Server port (8443)
ss -tlnp | grep 8443

# Check JupyterLab port (8888)
ss -tlnp | grep 8888

# Use curl to check services
curl -s -o /dev/null -w "%{http_code}" http://127.0.0.1:8443
curl -s -o /dev/null -w "%{http_code}" http://127.0.0.1:8888

Health Checks

Automated Health Check Script

#!/bin/bash
# vibe-health-check.sh

echo "=== VIBE Service Health Check ==="

# Code-Server
if systemctl is-active --quiet code-server; then
    echo "[OK] Code-Server: Running"
else
    echo "[ERROR] Code-Server: Not running"
fi

# JupyterLab
if systemctl is-active --quiet jupyter; then
    echo "[OK] JupyterLab: Running"
else
    echo "[ERROR] JupyterLab: Not running"
fi

# Port checks
if ss -tlnp | grep -q ":8443 "; then
    echo "[OK] Code-Server port 8443: Listening"
else
    echo "[ERROR] Code-Server port 8443: Not listening"
fi

if ss -tlnp | grep -q ":8888 "; then
    echo "[OK] JupyterLab port 8888: Listening"
else
    echo "[ERROR] JupyterLab port 8888: Not listening"
fi

Endpoint Checks with curl

# Check Code-Server
curl -s -o /dev/null -w "%{http_code}\n" \
    --connect-timeout 5 \
    http://127.0.0.1:8443/healthz

# Check JupyterLab
curl -s -o /dev/null -w "%{http_code}\n" \
    --connect-timeout 5 \
    http://127.0.0.1:8888/api/status

Alert Configuration

Service Down Alerts

Configure VIBE service alerts in Alertmanager:

# Example: Code-Server down alert
- alert: CodeServerDown
  expr: up{job="code-server"} == 0
  for: 5m
  labels:
    severity: warning
  annotations:
    summary: "Code-Server service down"
    description: "Code-Server on {{ $labels.instance }} has been down for over 5 minutes"

# Example: JupyterLab down alert
- alert: JupyterLabDown
  expr: up{job="jupyter"} == 0
  for: 5m
  labels:
    severity: warning
  annotations:
    summary: "JupyterLab service down"
    description: "JupyterLab on {{ $labels.instance }} has been down for over 5 minutes"

Log Management

Log Locations

ComponentLog Location
Code-Serverjournalctl -u code-server
JupyterLabjournalctl -u jupyter
Claude CodeVictoriaLogs / ~/.claude/logs/

Log Queries

Using journalctl

# View recent logs
journalctl -u code-server -n 100

# View specific time range
journalctl -u code-server --since "2024-01-01" --until "2024-01-02"

# Real-time follow
journalctl -u code-server -f

Query Claude logs via VictoriaLogs

# Query via Grafana Explore panel
# Or use vlogs API
curl -G 'http://localhost:9428/select/logsql/query' \
    --data-urlencode 'query={job="claude"}'

Resource Monitoring

View Resource Usage

# View Code-Server resource usage
systemctl status code-server
ps aux | grep code-server

# View JupyterLab resource usage
systemctl status jupyter
ps aux | grep jupyter

# View memory usage
free -h

# View disk usage
df -h /data
df -h /fs

JuiceFS Storage Monitoring

If using JuiceFS as working directory:

# View JuiceFS status
juicefs status /fs

# View filesystem usage
df -h /fs

# View file stats
juicefs info /fs

Performance Tuning

Code-Server Performance Optimization

# Increase file handle limit (already in systemd service)
# /etc/systemd/system/code-server.service
LimitNOFILE=65535

JupyterLab Performance Optimization

# /data/jupyter/jupyter_config.py

# Increase max buffer size
c.ServerApp.max_buffer_size = 1073741824  # 1GB

# Configure timeout
c.ServerApp.shutdown_no_activity_timeout = 0  # Disable auto-shutdown

Last Modified 2026-01-25: v4.0 batch update (65761a0)