mcptesta/scripts/health-check.sh

#!/bin/sh
# MCPTesta Documentation Health Check Script
# Comprehensive health validation for container monitoring

set -e

# Configuration
HOST=${HOST:-localhost}
PORT=${PORT:-4321}
TIMEOUT=${HEALTH_TIMEOUT:-10}
MAX_RESPONSE_TIME=5000  # milliseconds

# Colors for output (simplified for sh compatibility)
RED='\033[0;31m'
GREEN='\033[0;32m'
NC='\033[0m'

# Logging functions
log() {
    echo "[$(date +'%Y-%m-%d %H:%M:%S')] $1"
}

success() {
    echo "${GREEN}[HEALTHY]${NC} $1"
}

error() {
    echo "${RED}[UNHEALTHY]${NC} $1" >&2
    exit 1
}

# Health check functions
check_port() {
    if ! nc -z "$HOST" "$PORT" 2>/dev/null; then
        error "Port $PORT is not accessible on $HOST"
    fi
    log "Port $PORT is accessible"
}

check_http_response() {
    local response_code
    local response_time
    
    # Check HTTP response with timeout
    if ! response_code=$(wget --spider --server-response --timeout="$TIMEOUT" --tries=1 \
        "http://$HOST:$PORT/" 2>&1 | grep "HTTP/" | tail -1 | awk '{print $2}'); then
        error "HTTP request failed or timed out"
    fi
    
    # Validate response code
    if [ "$response_code" != "200" ]; then
        error "HTTP response code: $response_code (expected: 200)"
    fi
    
    log "HTTP response: $response_code OK"
}

check_response_time() {
    local start_time
    local end_time
    local response_time
    
    start_time=$(date +%s%3N)
    
    if ! wget --spider --quiet --timeout="$TIMEOUT" --tries=1 "http://$HOST:$PORT/" 2>/dev/null; then
        error "Response time check failed"
    fi
    
    end_time=$(date +%s%3N)
    response_time=$((end_time - start_time))
    
    if [ "$response_time" -gt "$MAX_RESPONSE_TIME" ]; then
        error "Response time too slow: ${response_time}ms (max: ${MAX_RESPONSE_TIME}ms)"
    fi
    
    log "Response time: ${response_time}ms"
}

check_content() {
    local content
    
    # Check if the page contains expected content
    if ! content=$(wget --quiet --timeout="$TIMEOUT" --tries=1 -O- "http://$HOST:$PORT/" 2>/dev/null); then
        error "Failed to retrieve page content"
    fi
    
    # Basic content validation
    if ! echo "$content" | grep -q "MCPTesta"; then
        error "Page content validation failed - 'MCPTesta' not found"
    fi
    
    if ! echo "$content" | grep -q "<html"; then
        error "Page content validation failed - HTML structure not found"
    fi
    
    log "Content validation passed"
}

check_dependencies() {
    # Check if required commands are available
    command -v wget >/dev/null 2>&1 || error "wget command not found"
    command -v nc >/dev/null 2>&1 || error "nc (netcat) command not found"
    
    log "Required dependencies available"
}

check_memory_usage() {
    local memory_usage
    local memory_limit_mb=512  # Default limit
    
    # Get memory usage in MB (simplified check)
    if [ -f /proc/meminfo ]; then
        memory_usage=$(awk '/MemAvailable/ {printf "%.0f", $2/1024}' /proc/meminfo)
        
        if [ "$memory_usage" -lt 50 ]; then
            error "Low available memory: ${memory_usage}MB"
        fi
        
        log "Available memory: ${memory_usage}MB"
    else
        log "Memory check skipped (no /proc/meminfo)"
    fi
}

check_disk_space() {
    local disk_usage
    local disk_limit=90  # 90% threshold
    
    # Check disk usage of /app
    if disk_usage=$(df /app 2>/dev/null | tail -1 | awk '{print $5}' | sed 's/%//'); then
        if [ "$disk_usage" -gt "$disk_limit" ]; then
            error "High disk usage: ${disk_usage}% (limit: ${disk_limit}%)"
        fi
        
        log "Disk usage: ${disk_usage}%"
    else
        log "Disk check skipped"
    fi
}

# Main health check routine
main() {
    log "Starting comprehensive health check..."
    log "Target: http://$HOST:$PORT/"
    log "Timeout: ${TIMEOUT}s"
    
    # Run all health checks
    check_dependencies
    check_memory_usage
    check_disk_space
    check_port
    check_http_response
    check_response_time
    check_content
    
    success "All health checks passed"
    log "Container is healthy and ready to serve requests"
}

# Run health check
main