proxmox/scripts/archive/consolidated/fix/fix-pve2-container-storage.sh

#!/usr/bin/env bash
# Fix missing storage volumes for containers on pve2
# This script checks storage config and recreates missing volumes
# Usage: ./scripts/fix-pve2-container-storage.sh [--dry-run]

set -euo pipefail

# Load IP configuration
SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
PROJECT_ROOT="$(cd "$SCRIPT_DIR/.." && pwd)"
source "${PROJECT_ROOT}/config/ip-addresses.conf" 2>/dev/null || true


PVE2_IP="${PROXMOX_HOST_R630_01}"
DRY_RUN=false

if [[ "${1:-}" == "--dry-run" ]]; then
    DRY_RUN=true
    echo "DRY RUN MODE - No changes will be made"
fi

# 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}[WARN]${NC} $1"; }
log_error() { echo -e "${RED}[ERROR]${NC} $1"; }

# All containers
ALL_CONTAINERS=(3000 3001 3002 3003 3500 3501 5200 6000 6400 10000 10001 10020 10030 10040 10050 10060 10070 10080 10090 10091 10092 10100 10101 10120 10130 10150 10151 10200 10201 10202 10210 10230 10232)

echo ""
log_info "Fixing container storage on pve2..."
echo ""

# Check SSH
if ! ssh -o ConnectTimeout=5 -o StrictHostKeyChecking=no root@${PVE2_IP} "echo 'OK'" &>/dev/null; then
    log_error "Cannot access pve2"
    exit 1
fi

FIXED=0
SKIPPED=0
FAILED=0

for vmid in "${ALL_CONTAINERS[@]}"; do
    log_info "Checking CT $vmid..."
    
    # Get storage config
    rootfs_config=$(ssh -o ConnectTimeout=5 -o StrictHostKeyChecking=no root@${PVE2_IP} \
        "pct config $vmid 2>/dev/null | grep '^rootfs:'" || echo "")
    
    if [[ -z "$rootfs_config" ]]; then
        log_warn "  Config missing, skipping..."
        ((SKIPPED++))
        continue
    fi
    
    # Extract storage pool and volume
    storage_line=$(echo "$rootfs_config" | sed 's/^rootfs: //')
    storage_pool=$(echo "$storage_line" | cut -d':' -f1)
    volume_name=$(echo "$storage_line" | cut -d':' -f2 | cut -d',' -f1)
    size=$(echo "$storage_line" | grep -oP 'size=\K[^,]+' || echo "")
    
    echo "  Storage: $storage_pool:$volume_name"
    if [[ -n "$size" ]]; then
        echo "  Size: $size"
    fi
    
    # Check if volume exists
    volume_exists=$(ssh -o ConnectTimeout=5 -o StrictHostKeyChecking=no root@${PVE2_IP} \
        "lvs 2>/dev/null | grep -q \"^${volume_name}\" && echo 'exists' || echo 'missing'" || echo "error")
    
    if [[ "$volume_exists" == "exists" ]]; then
        log_success "  ✓ Volume exists"
        # Try to start
        if [[ "$DRY_RUN" == "false" ]]; then
            if ssh -o ConnectTimeout=5 -o StrictHostKeyChecking=no root@${PVE2_IP} \
                "pct start $vmid" 2>&1 >/dev/null; then
                log_success "  ✓ Container started"
                ((FIXED++))
            else
                log_warn "  ⚠️  Volume exists but start failed"
                ((FAILED++))
            fi
        else
            log_info "  [DRY RUN] Would attempt to start"
        fi
    else
        log_warn "  ⚠️  Volume missing: $volume_name"
        
        # Check if storage pool is active
        pool_status=$(ssh -o ConnectTimeout=5 -o StrictHostKeyChecking=no root@${PVE2_IP} \
            "pvesm status 2>/dev/null | awk '\$1 == \"$storage_pool\" {print \$3}'" || echo "unknown")
        
        if [[ "$pool_status" != "active" ]]; then
            log_error "  ❌ Storage pool '$storage_pool' is not active (status: $pool_status)"
            log_info "  Checking for alternative storage..."
            
            # Find active storage
            alt_storage=$(ssh -o ConnectTimeout=5 -o StrictHostKeyChecking=no root@${PVE2_IP} \
                "pvesm status 2>/dev/null | grep -E 'active|available' | grep lvmthin | awk '{print \$1}' | head -1" || echo "")
            
            if [[ -n "$alt_storage" ]]; then
                log_info "  Found alternative: $alt_storage"
                if [[ "$DRY_RUN" == "false" ]] && [[ -n "$size" ]]; then
                    log_info "  Updating storage config..."
                    new_rootfs="${alt_storage}:${volume_name},size=${size}"
                    if ssh -o ConnectTimeout=5 -o StrictHostKeyChecking=no root@${PVE2_IP} \
                        "pct set $vmid -rootfs $new_rootfs" 2>&1 >/dev/null; then
                        log_success "  ✓ Storage config updated"
                        # Try to start (will create volume)
                        if ssh -o ConnectTimeout=5 -o StrictHostKeyChecking=no root@${PVE2_IP} \
                            "pct start $vmid" 2>&1 >/dev/null; then
                            log_success "  ✓ Container started"
                            ((FIXED++))
                        else
                            log_warn "  ⚠️  Start failed after config update"
                            ((FAILED++))
                        fi
                    else
                        log_error "  ❌ Failed to update storage config"
                        ((FAILED++))
                    fi
                else
                    log_info "  [DRY RUN] Would update to $alt_storage and start"
                fi
            else
                log_error "  ❌ No active storage found"
                ((FAILED++))
            fi
        else
            log_info "  Storage pool is active, volume needs to be created"
            log_info "  Attempting to start container (will create volume)..."
            
            if [[ "$DRY_RUN" == "false" ]]; then
                if ssh -o ConnectTimeout=5 -o StrictHostKeyChecking=no root@${PVE2_IP} \
                    "pct start $vmid" 2>&1 >/dev/null; then
                    log_success "  ✓ Container started (volume created)"
                    ((FIXED++))
                else
                    error=$(ssh -o ConnectTimeout=5 -o StrictHostKeyChecking=no root@${PVE2_IP} \
                        "pct start $vmid 2>&1" || true)
                    log_error "  ❌ Start failed:"
                    echo "$error" | sed 's/^/    /' | head -3
                    ((FAILED++))
                fi
            else
                log_info "  [DRY RUN] Would attempt: pct start $vmid"
            fi
        fi
    fi
    
    echo ""
done

echo ""
log_info "Summary:"
log_info "  Fixed: $FIXED"
log_info "  Failed: $FAILED"
log_info "  Skipped: $SKIPPED"

if [[ "$DRY_RUN" == "true" ]]; then
    log_warn "This was a DRY RUN - no changes were made"
fi