proxmox/scripts/monitoring/monitor-blockchain-health.sh.bak

#!/bin/bash
# Comprehensive Blockchain Health Monitoring Script
# Monitors block production, transaction inclusion, and node health

set -euo pipefail

RPC_URL="${RPC_URL:-http://192.168.11.211:8545}"
DEPLOYER="${DEPLOYER:-0x4A666F96fC8764181194447A7dFdb7d471b301C8}"
PROXMOX_USER="${PROXMOX_USER:-root}"
PROXMOX_ML110="${PROXMOX_ML110:-192.168.11.10}"
PROXMOX_R630="${PROXMOX_R630:-192.168.11.11}"

# Colors
RED='\033[0;31m'
GREEN='\033[0;32m'
YELLOW='\033[1;33m'
BLUE='\033[0;34m'
CYAN='\033[0;36m'
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} $1"; }
log_section() { echo -e "\n${CYAN}━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━${NC}"; echo -e "${CYAN}$1${NC}"; echo -e "${CYAN}━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━${NC}\n"; }

echo "=== Blockchain Health Monitor ==="
echo "Timestamp: $(date '+%Y-%m-%d %H:%M:%S')"
echo ""

# Check RPC connectivity
log_section "RPC Node Status"
if timeout 5 cast chain-id --rpc-url "$RPC_URL" >/dev/null 2>&1; then
    CHAIN_ID=$(cast chain-id --rpc-url "$RPC_URL" 2>/dev/null)
    BLOCK_NUM=$(cast block-number --rpc-url "$RPC_URL" 2>/dev/null)
    BLOCK_DEC=$(cast --to-dec "$BLOCK_NUM" 2>/dev/null || echo "0")
    log_success "RPC accessible"
    echo "  Chain ID: $CHAIN_ID"
    echo "  Latest block: $BLOCK_DEC ($BLOCK_NUM)"
else
    log_error "RPC not accessible"
    exit 1
fi

# Check block production
log_section "Block Production"
BLOCK1=$(cast block-number --rpc-url "$RPC_URL" 2>/dev/null)
sleep 5
BLOCK2=$(cast block-number --rpc-url "$RPC_URL" 2>/dev/null)
BLOCK1_DEC=$(cast --to-dec "$BLOCK1" 2>/dev/null || echo "0")
BLOCK2_DEC=$(cast --to-dec "$BLOCK2" 2>/dev/null || echo "0")
BLOCK_DIFF=$((BLOCK2_DEC - BLOCK1_DEC))

if [ "$BLOCK_DIFF" -gt 0 ]; then
    log_success "Blocks being produced ($BLOCK_DIFF blocks in 5s)"
else
    log_error "Block production stalled (no new blocks in 5s)"
fi

# Check transaction inclusion
log_section "Transaction Inclusion"
TX_COUNT_TOTAL=0
EMPTY_BLOCKS=0
for i in 0 1 2 3 4 5; do
    BLOCK_NUM=$((BLOCK2_DEC - i))
    BLOCK_HEX=$(printf '0x%x' $BLOCK_NUM)
    TX_COUNT=$(cast rpc eth_getBlockTransactionCountByNumber "$BLOCK_HEX" --rpc-url "$RPC_URL" 2>/dev/null | tr -d '"')
    TX_COUNT_DEC=$(cast --to-dec "$TX_COUNT" 2>/dev/null || echo "0")
    TX_COUNT_TOTAL=$((TX_COUNT_TOTAL + TX_COUNT_DEC))
    if [ "$TX_COUNT_DEC" -eq 0 ]; then
        EMPTY_BLOCKS=$((EMPTY_BLOCKS + 1))
    fi
done

if [ "$TX_COUNT_TOTAL" -gt 0 ]; then
    log_success "Transactions being included ($TX_COUNT_TOTAL txs in last 6 blocks)"
else
    log_warn "No transactions in last 6 blocks ($EMPTY_BLOCKS empty blocks)"
fi

# Check pending transactions
log_section "Pending Transactions"
LATEST_HEX=$(cast rpc eth_getTransactionCount "$DEPLOYER" latest --rpc-url "$RPC_URL" 2>/dev/null | tr -d '"')
PENDING_HEX=$(cast rpc eth_getTransactionCount "$DEPLOYER" pending --rpc-url "$RPC_URL" 2>/dev/null | tr -d '"')
LATEST_DEC=$(cast --to-dec "$LATEST_HEX" 2>/dev/null || echo "0")
PENDING_DEC=$(cast --to-dec "$PENDING_HEX" 2>/dev/null || echo "0")
PENDING_COUNT=$((PENDING_DEC - LATEST_DEC))

if [ "$PENDING_COUNT" -eq 0 ]; then
    log_success "No pending transactions"
else
    log_warn "$PENDING_COUNT pending transactions (nonces $((LATEST_DEC + 1))-$PENDING_DEC)"
fi

# Check validator status
log_section "Validator Status"
VALIDATORS=(
    "1000:$PROXMOX_R630"
    "1001:$PROXMOX_R630"
    "1002:$PROXMOX_R630"
    "1003:$PROXMOX_ML110"
    "1004:$PROXMOX_ML110"
)

ACTIVE_COUNT=0
for validator in "${VALIDATORS[@]}"; do
    IFS=':' read -r VMID HOST <<< "$validator"
    SSH_TARGET="${PROXMOX_USER}@${HOST}"
    STATUS=$(ssh -o ConnectTimeout=5 -o StrictHostKeyChecking=no "$SSH_TARGET" \
        "pct exec $VMID -- systemctl is-active besu-validator" 2>/dev/null || echo "unknown")
    
    if [ "$STATUS" = "active" ]; then
        ACTIVE_COUNT=$((ACTIVE_COUNT + 1))
        echo "  Validator $VMID: $STATUS"
    else
        log_warn "Validator $VMID: $STATUS"
    fi
done

if [ "$ACTIVE_COUNT" -eq 5 ]; then
    log_success "All 5 validators active"
else
    log_error "Only $ACTIVE_COUNT/5 validators active"
fi

# Check peer connections
log_section "Peer Connections"
PEER_COUNT=$(cast rpc admin_peers --rpc-url "$RPC_URL" 2>/dev/null | jq '. | length' 2>/dev/null || echo "N/A")
if [ "$PEER_COUNT" != "N/A" ] && [ "$PEER_COUNT" -ge 5 ]; then
    log_success "RPC has $PEER_COUNT peer connections"
else
    log_warn "RPC has $PEER_COUNT peer connections (expected >= 5)"
fi

# Summary
log_section "Health Summary"
ISSUES=0

if [ "$BLOCK_DIFF" -eq 0 ]; then
    log_error "❌ Block production stalled"
    ISSUES=$((ISSUES + 1))
else
    log_success "✓ Block production active"
fi

if [ "$TX_COUNT_TOTAL" -eq 0 ] && [ "$PENDING_COUNT" -gt 0 ]; then
    log_error "❌ Transactions not being included"
    ISSUES=$((ISSUES + 1))
elif [ "$TX_COUNT_TOTAL" -gt 0 ]; then
    log_success "✓ Transactions being included"
fi

if [ "$ACTIVE_COUNT" -lt 5 ]; then
    log_error "❌ Not all validators active"
    ISSUES=$((ISSUES + 1))
else
    log_success "✓ All validators active"
fi

if [ "$PENDING_COUNT" -gt 10 ]; then
    log_warn "⚠ High number of pending transactions ($PENDING_COUNT)"
    ISSUES=$((ISSUES + 1))
fi

echo ""
if [ "$ISSUES" -eq 0 ]; then
    log_success "Overall Status: HEALTHY"
    exit 0
else
    log_error "Overall Status: $ISSUES issue(s) detected"
    exit 1
fi