proxmox/scripts/fix-ssl-certificate-error-596.sh

#!/bin/bash
# Fix Proxmox SSL Certificate Error 596
# Error: error:0A000086:SSL routines::certificate verify failed (596)
# Usage: ./scripts/fix-ssl-certificate-error-596.sh [node_ip|all]

set -euo pipefail

SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
PROJECT_ROOT="$(cd "$SCRIPT_DIR/.." && pwd)"

# Colors
RED='\033[0;31m'
GREEN='\033[0;32m'
YELLOW='\033[1;33m'
BLUE='\033[0;34m'
NC='\033[0m'

log_info() { echo -e "${BLUE}[INFO]${NC} $1"; }
log_success() { echo -e "${GREEN}[✓]${NC} $1"; }
log_warn() { echo -e "${YELLOW}[⚠]${NC} $1"; }
log_error() { echo -e "${RED}[✗]${NC} $1"; }

# Proxmox cluster nodes
declare -A NODES
NODES[ml110]="192.168.11.10"
NODES[r630-01]="192.168.11.11"
NODES[r630-02]="192.168.11.12"
NODES[r630-03]="192.168.11.13"
NODES[r630-04]="192.168.11.14"

fix_node() {
    local node_ip="$1"
    local node_name="${2:-$node_ip}"
    
    log_info "=== Fixing SSL certificates on ${node_name} (${node_ip}) ==="
    echo ""
    
    # Test connectivity
    if ! ping -c 2 -W 2 "$node_ip" >/dev/null 2>&1; then
        log_error "Node ${node_ip} is NOT reachable"
        return 1
    fi
    
    log_info "Connecting to ${node_ip}..."
    echo ""
    
    # Check if we can SSH without password (key-based auth)
    if ssh -o ConnectTimeout=5 -o StrictHostKeyChecking=no -o BatchMode=yes root@"$node_ip" "echo 'Connected'" >/dev/null 2>&1; then
        log_info "Using SSH key authentication"
        SSH_CMD="ssh -o StrictHostKeyChecking=no root@${node_ip}"
    else
        log_warn "SSH key authentication not available"
        log_info "You will need to run the commands manually on the node:"
        echo ""
        echo "SSH to the node:"
        echo "  ssh root@${node_ip}"
        echo ""
        echo "Then run:"
        echo "  pvecm updatecerts -f"
        echo "  systemctl restart pveproxy pvedaemon"
        echo ""
        return 1
    fi
    
    log_info "Executing SSL certificate fix..."
    echo ""
    
    $SSH_CMD bash <<'ENDSSH'
        set -e
        
        echo "=== Step 1: Checking current certificate status ==="
        if [ -f /etc/pve/pve-root-ca.pem ]; then
            echo "Current certificate dates:"
            openssl x509 -in /etc/pve/pve-root-ca.pem -noout -dates 2>/dev/null || echo "Could not read certificate dates"
        else
            echo "Certificate file not found (may be normal)"
        fi
        echo ""
        
        echo "=== Step 2: Regenerating SSL certificates ==="
        if command -v pvecm >/dev/null 2>&1; then
            pvecm updatecerts -f
            echo "✓ Certificates regenerated"
        else
            echo "ERROR: pvecm command not found"
            exit 1
        fi
        echo ""
        
        echo "=== Step 3: Restarting Proxmox services ==="
        systemctl restart pveproxy pvedaemon
        sleep 2
        echo "✓ Services restarted"
        echo ""
        
        echo "=== Step 4: Verifying services are running ==="
        if systemctl is-active --quiet pveproxy && systemctl is-active --quiet pvedaemon; then
            echo "✓ pveproxy: active"
            echo "✓ pvedaemon: active"
        else
            echo "⚠ Some services may not be running properly"
            systemctl status pveproxy --no-pager -l | head -5 || true
            systemctl status pvedaemon --no-pager -l | head -5 || true
        fi
        echo ""
        
        echo "=== Step 5: Verifying new certificate ==="
        if [ -f /etc/pve/pve-root-ca.pem ]; then
            echo "New certificate dates:"
            openssl x509 -in /etc/pve/pve-root-ca.pem -noout -dates 2>/dev/null || echo "Could not read certificate dates"
        fi
        echo ""
        
        echo "=== Step 6: Testing web interface ==="
        if curl -k -s -o /dev/null -w "HTTP Status: %{http_code}\n" https://localhost:8006/ >/dev/null 2>&1; then
            HTTP_CODE=$(curl -k -s -o /dev/null -w "%{http_code}" https://localhost:8006/ 2>/dev/null || echo "000")
            if [ "$HTTP_CODE" = "200" ] || [ "$HTTP_CODE" = "401" ] || [ "$HTTP_CODE" = "302" ]; then
                echo "✓ Web interface is responding (HTTP $HTTP_CODE)"
            else
                echo "⚠ Web interface returned HTTP $HTTP_CODE"
            fi
        else
            echo "⚠ Could not test web interface"
        fi
        echo ""
ENDSSH
    
    if [ $? -eq 0 ]; then
        log_success "SSL certificate fix completed for ${node_name}"
        
        # Test from remote
        log_info "Testing web interface from remote..."
        sleep 2
        HTTP_CODE=$(curl -k -s -o /dev/null -w "%{http_code}" --connect-timeout 5 "https://${node_ip}:8006/" 2>/dev/null || echo "000")
        if [ "$HTTP_CODE" = "200" ] || [ "$HTTP_CODE" = "401" ] || [ "$HTTP_CODE" = "302" ]; then
            log_success "Web interface is accessible at https://${node_ip}:8006"
        else
            log_warn "Web interface test returned HTTP $HTTP_CODE"
            log_info "You may need to clear your browser cache and cookies"
        fi
    else
        log_error "SSL certificate fix failed for ${node_name}"
        return 1
    fi
    
    echo ""
    echo "----------------------------------------"
    echo ""
}

# Determine target
TARGET="${1:-all}"

if [[ "$TARGET" == "all" ]]; then
    log_info "Fixing SSL certificates on all Proxmox nodes..."
    echo ""
    
    for node_name in "${!NODES[@]}"; do
        node_ip="${NODES[$node_name]}"
        fix_node "$node_ip" "$node_name" || log_warn "Failed to fix ${node_name}, continuing..."
    done
    
    log_success "All fix attempts complete!"
    echo ""
    log_info "Next steps:"
    log_info "  1. Clear browser cache and cookies"
    log_info "  2. Access Proxmox UI: https://<node-ip>:8006"
    log_info "  3. Accept certificate warning if prompted (first time only)"
    
elif [[ -n "${NODES[$TARGET]:-}" ]]; then
    # Target is a node name
    node_ip="${NODES[$TARGET]}"
    fix_node "$node_ip" "$TARGET"
    
elif [[ "$TARGET" =~ ^[0-9]+\.[0-9]+\.[0-9]+\.[0-9]+$ ]]; then
    # Target is an IP address
    fix_node "$TARGET" "$TARGET"
    
else
    log_error "Invalid target: $TARGET"
    echo ""
    echo "Usage: $0 [node_name|node_ip|all]"
    echo ""
    echo "Available nodes:"
    for node_name in "${!NODES[@]}"; do
        echo "  - $node_name (${NODES[$node_name]})"
    done
    echo "  - all (fix all nodes)"
    exit 1
fi