155 lines
5.4 KiB
Bash
155 lines
5.4 KiB
Bash
#!/usr/bin/env bash
|
|
# Fix all validators: remove legacy tx-pool options, normalize validator networking,
|
|
# and restart besu-validator. Run from project root (sources config/ip-addresses.conf).
|
|
#
|
|
# Usage: bash scripts/fix-all-validators-and-txpool.sh [--dry-run]
|
|
|
|
set -euo pipefail
|
|
|
|
SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
|
|
PROJECT_ROOT="$(cd "$SCRIPT_DIR/.." && pwd)"
|
|
cd "$PROJECT_ROOT"
|
|
[ -f config/ip-addresses.conf ] && source config/ip-addresses.conf 2>/dev/null || true
|
|
[ -f scripts/lib/load-project-env.sh ] && source scripts/lib/load-project-env.sh 2>/dev/null || true
|
|
|
|
PROXMOX_SSH_USER="${PROXMOX_SSH_USER:-root}"
|
|
R630_01="${PROXMOX_HOST_R630_01:-${PROXMOX_R630_01:-192.168.11.11}}"
|
|
|
|
validator_host() {
|
|
local vmid="$1"
|
|
if type get_host_for_vmid >/dev/null 2>&1; then
|
|
get_host_for_vmid "$vmid"
|
|
elif [[ "$vmid" -le 1002 ]]; then
|
|
echo "$R630_01"
|
|
else
|
|
echo "${PROXMOX_HOST_ML110:-192.168.11.10}"
|
|
fi
|
|
}
|
|
|
|
validator_ip() {
|
|
local vmid="$1"
|
|
case "$vmid" in
|
|
1000) echo "${IP_VALIDATOR_0:-192.168.11.100}" ;;
|
|
1001) echo "${IP_VALIDATOR_1:-192.168.11.101}" ;;
|
|
1002) echo "${IP_VALIDATOR_2:-192.168.11.102}" ;;
|
|
1003) echo "${IP_VALIDATOR_3:-192.168.11.103}" ;;
|
|
1004) echo "${IP_VALIDATOR_4:-192.168.11.104}" ;;
|
|
*) return 1 ;;
|
|
esac
|
|
}
|
|
|
|
VALIDATORS=(
|
|
"1000:$(validator_host 1000)"
|
|
"1001:$(validator_host 1001)"
|
|
"1002:$(validator_host 1002)"
|
|
"1003:$(validator_host 1003)"
|
|
"1004:$(validator_host 1004)"
|
|
)
|
|
|
|
CONFIG_PATH="/etc/besu/config-validator.toml"
|
|
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} $1"; }
|
|
|
|
DRY_RUN=false
|
|
[[ "${1:-}" = "--dry-run" ]] && DRY_RUN=true
|
|
|
|
echo ""
|
|
echo "=== Fix All Validators + Tx-Pool (Evict Stuck) ==="
|
|
echo ""
|
|
|
|
fix_one() {
|
|
local vmid="$1"
|
|
local host="$2"
|
|
local validator_ip
|
|
local ssh_target="${PROXMOX_SSH_USER}@${host}"
|
|
|
|
validator_ip="$(validator_ip "$vmid")" || {
|
|
log_error " Could not determine validator IP for VMID $vmid"
|
|
return 1
|
|
}
|
|
|
|
log_info "Validator $vmid on $host"
|
|
|
|
# Check container running
|
|
local status
|
|
status=$(ssh -o ConnectTimeout=5 -o StrictHostKeyChecking=no "$ssh_target" "pct status $vmid 2>/dev/null" | awk '{print $2}' || echo "unknown")
|
|
if [[ "$status" != "running" ]]; then
|
|
log_warn " VMID $vmid not running (status: $status) — skip"
|
|
return 0
|
|
fi
|
|
|
|
if "$DRY_RUN"; then
|
|
log_info " [dry-run] Would remove legacy tx-pool lines, set p2p-host=${validator_ip}, force sync-mode=FULL, and restart besu-validator"
|
|
return 0
|
|
fi
|
|
|
|
# Remove legacy options (cause crash with layered pool); normalize validator networking.
|
|
ssh -o ConnectTimeout=5 -o StrictHostKeyChecking=no "$ssh_target" "pct exec $vmid -- bash -c '
|
|
set -e
|
|
CFG=/etc/besu/config-validator.toml
|
|
[ -f /config/config-validator.toml ] && CFG=/config/config-validator.toml
|
|
if [ ! -f \"\$CFG\" ]; then echo \"Config not found: \$CFG\"; exit 1; fi
|
|
sed -i \"/^tx-pool-max-size=/d\" \"\$CFG\" 2>/dev/null || true
|
|
sed -i \"/^tx-pool-limit-by-account-percentage=/d\" \"\$CFG\" 2>/dev/null || true
|
|
sed -i \"/^tx-pool-retention-hours=/d\" \"\$CFG\" 2>/dev/null || true
|
|
sed -i \"s|^p2p-host=.*|p2p-host=\\\"${validator_ip}\\\"|\" \"\$CFG\"
|
|
sed -i \"s|^sync-mode=.*|sync-mode=\\\"FULL\\\"|\" \"\$CFG\"
|
|
grep -q \"^p2p-host=\" \"\$CFG\" || echo \"p2p-host=\\\"${validator_ip}\\\"\" >> \"\$CFG\"
|
|
grep -q \"^sync-mode=\" \"\$CFG\" || echo \"sync-mode=\\\"FULL\\\"\" >> \"\$CFG\"
|
|
if ! grep -q \"tx-pool-max-future-by-sender\" \"\$CFG\"; then
|
|
echo \"\" >> \"\$CFG\"
|
|
echo \"# Layered Transaction Pool (Besu 23.10+; keep future queue tight)\" >> \"\$CFG\"
|
|
echo \"tx-pool-max-future-by-sender=1\" >> \"\$CFG\"
|
|
echo \"tx-pool-layer-max-capacity=12500000\" >> \"\$CFG\"
|
|
echo \"tx-pool-max-prioritized=2000\" >> \"\$CFG\"
|
|
echo \"tx-pool-price-bump=10\" >> \"\$CFG\"
|
|
else
|
|
sed -i \"s/^tx-pool-max-future-by-sender=.*/tx-pool-max-future-by-sender=1/\" \"\$CFG\"
|
|
fi
|
|
# tx-pool-min-score=0 not added: unsupported in some Besu builds (causes Unknown option and crash loop)
|
|
sed -i \"/^tx-pool-min-score=/d\" \"\$CFG\" 2>/dev/null || true
|
|
'" 2>/dev/null || { log_error " SSH/config update failed"; return 1; }
|
|
|
|
log_success " Config updated"
|
|
|
|
# Restart with a hard fallback so a wedged Besu process cannot block the whole recovery.
|
|
if ! ssh -o ConnectTimeout=5 -o StrictHostKeyChecking=no "$ssh_target" "pct exec $vmid -- bash -lc '
|
|
timeout 30 systemctl restart besu-validator || {
|
|
systemctl kill -s SIGKILL besu-validator || true
|
|
sleep 2
|
|
systemctl reset-failed besu-validator || true
|
|
systemctl start besu-validator
|
|
}
|
|
'" 2>/dev/null; then
|
|
log_error " Restart failed"
|
|
return 1
|
|
fi
|
|
log_success " besu-validator restarted"
|
|
return 0
|
|
}
|
|
|
|
FAILED=0
|
|
for entry in "${VALIDATORS[@]}"; do
|
|
IFS=: read -r vmid host <<< "$entry"
|
|
if ! fix_one "$vmid" "$host"; then
|
|
((FAILED++)) || true
|
|
fi
|
|
echo ""
|
|
done
|
|
|
|
echo "=== Summary ==="
|
|
if [[ $FAILED -eq 0 ]]; then
|
|
log_success "All validators updated and restarted (or skipped if not running)."
|
|
log_info "Block production should resume once quorum is met. Check: bash scripts/monitoring/monitor-blockchain-health.sh"
|
|
else
|
|
log_error "$FAILED validator(s) failed. Check SSH and container status."
|
|
exit 1
|
|
fi
|