Pulsechain GETH watchdog

watchdog.sh

#!/bin/bash

# Notes:
# Figure out your restart command. The valuesbelow are mine and work for my docker compose setup. If've you installed native
# or using vanilla docker yours might be different. 
# After saving the file locally run: chmod -x <path-to-file>/watchdog.sh
# then run with <path-to-file>/watchdog.sh or if in the same folder then ./watchdog.sh

# ==========================================
# CONFIGURATION
# ==========================================
NODE_IP="192.168.50.103" # Update with your node's IP address, 127.0.0.1, 0.0.0.0, actual IP - whatever works for your setup
NODE_PORT="8547"         # default 8545
CHECK_INTERVAL=30        # Seconds between checks
TIMEOUT_LIMIT=120        # Seconds to wait for increase before restart (2 minutes)

# RESTART COMMAND
# RESTART_CMD="sudo systemctl restart geth"
# RESTART_CMD="docker restart geth"
RESTART_CMD="cd /home/hamish/Docker/geth-prysm && docker-compose down && docker-compose up -d" # use the name of the service, not the value in the name: field

# ==========================================
# INTERNAL LOGIC
# ==========================================
last_height=0
last_change_time=$(date +%s)

get_height() {
    local response=$(curl -s -X POST -H "Content-Type: application/json" --data '{"jsonrpc":"2.0","method":"eth_blockNumber","params":[],"id":1}' "http://$NODE_IP:$NODE_PORT")
    local hex=$(echo "$response" | grep -o '"result":"0x[0-9a-fA-F]*"' | sed 's/"result":"//;s/"//')
    [[ -z "$hex" ]] && echo 0 || echo $((hex))
}

echo "Geth Watchdog Active: http://$NODE_IP:$NODE_PORT"

while true; do
    height=$(get_height)
    now=$(date +%s)

    if [[ "$height" -gt "$last_height" ]]; then
        last_height=$height
        last_change_time=$now
        echo "[$(date +%T)] Height: $height"
    elif [[ $((now - last_change_time)) -ge "$TIMEOUT_LIMIT" ]]; then
        echo "[$(date +%T)] STALL DETECTED ($((now - last_change_time))s). Restarting..."
        eval "$RESTART_CMD"
        echo "[$(date +%T)] Restart triggered. Waiting 120s for node to initialize..."
        sleep 120
        last_change_time=$(date +%s)
        last_height=0
    else
        echo "[$(date +%T)] Stalled... $((now - last_change_time))s / $TIMEOUT_LIMIT s"
    fi

    sleep "$CHECK_INTERVAL"
done