infra-dashboard/app.py

#!/usr/bin/env python3
import os, json, time, logging
from datetime import datetime
from concurrent.futures import ThreadPoolExecutor, as_completed
from threading import Lock
from flask import Flask, render_template, jsonify
import paramiko
import urllib3
urllib3.disable_warnings(urllib3.exceptions.InsecureRequestWarning)

logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(levelname)s - %(message)s')
logger = logging.getLogger(__name__)
app = Flask(__name__)

CONFIG = {
    "refresh_interval": int(os.getenv("REFRESH_INTERVAL", "60")),
    "ssh_timeout": int(os.getenv("SSH_TIMEOUT", "10")),
    "ssh_key_path": os.getenv("SSH_KEY_PATH", "/app/ssh/id_ed25519"),
}

CONFIG_DIR = os.getenv("CONFIG_DIR", "/app/config")

def load_json_config(filename, default):
    for path in [os.path.join(CONFIG_DIR, filename), os.path.join(os.path.dirname(__file__), filename)]:
        if os.path.exists(path):
            with open(path) as f:
                return json.load(f)
    logger.warning(f"{filename} not found, using defaults")
    return default

hosts_config = load_json_config("hosts.json", {"proxmox_nodes": {}, "docker_hosts": {}})
PROXMOX_NODES = hosts_config["proxmox_nodes"]
DOCKER_HOSTS = hosts_config["docker_hosts"]

DIAGRAM_DATA = load_json_config("diagram.json", {})

class DataCache:
    def __init__(self):
        self.data = {}
        self.lock = Lock()
    def set(self, key, value):
        with self.lock: self.data[key] = value
    def get(self, key, default=None):
        with self.lock: return self.data.get(key, default)

cache = DataCache()

def get_ssh_client(host, user):
    client = paramiko.SSHClient()
    client.set_missing_host_key_policy(paramiko.AutoAddPolicy())
    pkey = None
    key_path = CONFIG["ssh_key_path"]
    if os.path.exists(key_path):
        try: pkey = paramiko.Ed25519Key.from_private_key_file(key_path)
        except:
            try: pkey = paramiko.RSAKey.from_private_key_file(key_path)
            except: pass
    client.connect(hostname=host, username=user, pkey=pkey, timeout=CONFIG["ssh_timeout"], allow_agent=True, look_for_keys=True)
    return client

def get_docker_containers(hostname, host_config):
    try:
        client = get_ssh_client(host_config["ip"], host_config["user"])
        cmd = "docker ps --format '{{.Names}}|{{.Status}}|{{.Image}}' 2>/dev/null"
        stdin, stdout, stderr = client.exec_command(cmd, timeout=15)
        output = stdout.read().decode('utf-8').strip()
        containers = []
        for line in output.split('\n'):
            if line and '|' in line:
                parts = line.split('|')
                if len(parts) >= 3:
                    status = parts[1].lower()
                    health = 'healthy' if 'healthy' in status else 'unhealthy' if 'unhealthy' in status else 'running' if 'up' in status else 'unknown'
                    containers.append({"name": parts[0], "status": parts[1], "image": parts[2], "health": health})
        client.close()
        return {"hostname": hostname, "status": "online", "container_count": len(containers), "containers": containers,
                "healthy": sum(1 for c in containers if c["health"] in ["healthy", "running"]),
                "unhealthy": sum(1 for c in containers if c["health"] == "unhealthy")}
    except Exception as e:
        logger.warning(f"Failed to connect to {hostname}: {e}")
        return {"hostname": hostname, "status": "offline", "container_count": 0, "containers": [], "error": str(e)}

def collect_all_hosts():
    results = {}
    with ThreadPoolExecutor(max_workers=10) as executor:
        futures = {executor.submit(get_docker_containers, name, config): name for name, config in DOCKER_HOSTS.items()}
        for future in as_completed(futures):
            hostname = futures[future]
            try: results[hostname] = future.result()
            except Exception as e: results[hostname] = {"hostname": hostname, "status": "error", "error": str(e)}
    return results

def refresh_data():
    logger.info("Refreshing data...")
    hosts = collect_all_hosts()
    cache.set("hosts", hosts)
    total = sum(h.get("container_count", 0) for h in hosts.values())
    online = sum(1 for h in hosts.values() if h.get("status") == "online")
    cache.set("summary", {"total_containers": total, "total_hosts": len(DOCKER_HOSTS), "online_hosts": online, "proxmox_nodes": len(PROXMOX_NODES)})
    cache.set("last_update", datetime.now().isoformat())
    logger.info(f"Refresh complete - {total} containers across {online}/{len(DOCKER_HOSTS)} hosts")

@app.route("/")
def index():
    return render_template("index.html", refresh_interval=CONFIG["refresh_interval"], docker_hosts=DOCKER_HOSTS, proxmox_nodes=PROXMOX_NODES)

@app.route("/api/data")
def api_data():
    return jsonify({"timestamp": cache.get("last_update"), "summary": cache.get("summary", {}), "hosts": cache.get("hosts", {}),
                    "config": {"docker_hosts": DOCKER_HOSTS, "proxmox_nodes": PROXMOX_NODES}})

@app.route("/api/refresh", methods=["POST"])
def api_refresh():
    refresh_data()
    return jsonify({"status": "ok", "timestamp": cache.get("last_update")})

@app.route("/api/diagram")
def api_diagram():
    return jsonify(DIAGRAM_DATA)

@app.route("/health")
def health():
    return jsonify({"status": "healthy", "last_update": cache.get("last_update")})

def background_refresh():
    while True:
        try: refresh_data()
        except Exception as e: logger.error(f"Refresh error: {e}")
        time.sleep(CONFIG["refresh_interval"])

import threading
refresh_data()
threading.Thread(target=background_refresh, daemon=True).start()

if __name__ == "__main__":
    app.run(host="0.0.0.0", port=5000, debug=False)