feat(manager): add Nexus Instance Manager for demo orchestration

- Complete Flask-based control panel for multi-tenant POS instances
- Dashboard with global stats, system health, and recent demos
- Demo provisioning in 1 click with auto-expiration tracking
- Tenant management: activate/deactivate, reset data, delete
- Health monitoring: PostgreSQL, Redis, disk, memory, systemd services
- Migration orchestration UI for running schema updates across all tenants
- JWT authentication with manager_users table
- Dark theme SPA frontend with real-time search and actions
- systemd service file included
This commit is contained in:
2026-05-17 21:01:01 +00:00
parent da362e32a6
commit be4bb8d9ad
20 changed files with 2685 additions and 0 deletions

View File

View File

@@ -0,0 +1,166 @@
"""Health monitoring service for Nexus infrastructure."""
import subprocess
import shutil
import socket
import urllib.request
import urllib.error
import psycopg2
import redis
from config import (
MASTER_DB_URL, REDIS_URL, POS_URL, DASHBOARD_URL, QUART_URL,
TENANT_DB_URL_TEMPLATE
)
def check_postgresql():
"""Check PostgreSQL connectivity."""
try:
conn = psycopg2.connect(MASTER_DB_URL, connect_timeout=5)
cur = conn.cursor()
cur.execute("SELECT version(), pg_database_size('nexus_autoparts')")
version, size = cur.fetchone()
cur.close()
conn.close()
return {
"status": "ok",
"version": version.split()[1] if version else "unknown",
"master_size_mb": round(size / (1024 * 1024), 2)
}
except Exception as e:
return {"status": "error", "error": str(e)}
def check_redis():
"""Check Redis connectivity."""
try:
r = redis.from_url(REDIS_URL, socket_connect_timeout=3)
info = r.info()
return {
"status": "ok",
"version": info.get("redis_version", "unknown"),
"used_memory_human": info.get("used_memory_human", "?"),
"connected_clients": info.get("connected_clients", 0)
}
except Exception as e:
return {"status": "error", "error": str(e)}
def check_http_service(name, url, timeout=5):
"""Generic HTTP health check."""
try:
req = urllib.request.Request(url, method="GET")
req.add_header("User-Agent", "Nexus-Manager/1.0")
with urllib.request.urlopen(req, timeout=timeout) as resp:
return {
"status": "ok",
"http_status": resp.status,
"latency_ms": None # Could add timing later
}
except urllib.error.HTTPError as e:
return {"status": "warning", "http_status": e.code, "error": str(e)}
except Exception as e:
return {"status": "error", "error": str(e)}
def check_disk_space(path="/"):
"""Check disk usage."""
try:
total, used, free = shutil.disk_usage(path)
return {
"status": "ok",
"total_gb": round(total / (1024**3), 2),
"used_gb": round(used / (1024**3), 2),
"free_gb": round(free / (1024**3), 2),
"percent_used": round((used / total) * 100, 1)
}
except Exception as e:
return {"status": "error", "error": str(e)}
def check_memory():
"""Check system memory via /proc/meminfo."""
try:
with open("/proc/meminfo") as f:
meminfo = f.read()
data = {}
for line in meminfo.splitlines():
if ":" in line:
key, value = line.split(":", 1)
data[key.strip()] = int(value.strip().split()[0]) # kB
total = data.get("MemTotal", 0) / 1024 / 1024 # GB
available = data.get("MemAvailable", data.get("MemFree", 0)) / 1024 / 1024
used = total - available
return {
"status": "ok",
"total_gb": round(total, 2),
"used_gb": round(used, 2),
"available_gb": round(available, 2),
"percent_used": round((used / total) * 100, 1) if total else 0
}
except Exception as e:
return {"status": "error", "error": str(e)}
def check_systemd_service(service_name):
"""Check systemd service status."""
try:
result = subprocess.run(
["systemctl", "is-active", service_name],
capture_output=True, text=True, timeout=5
)
active = result.stdout.strip() == "active"
return {
"status": "ok" if active else "warning",
"active": active,
"state": result.stdout.strip()
}
except Exception as e:
return {"status": "error", "error": str(e)}
def get_full_health_report():
"""Aggregate health report for all services."""
return {
"postgresql": check_postgresql(),
"redis": check_redis(),
"pos": check_http_service("pos", POS_URL),
"dashboard": check_http_service("dashboard", DASHBOARD_URL),
"quart": check_http_service("quart", QUART_URL),
"disk": check_disk_space(),
"memory": check_memory(),
"services": {
"nexus": check_systemd_service("nexus.service"),
"nexus-pos": check_systemd_service("nexus-pos.service"),
"nexus-quart": check_systemd_service("nexus-quart.service"),
"nexus-celery": check_systemd_service("nexus-celery.service"),
}
}
def get_tenant_health(db_name, timeout=5):
"""Check connectivity to a specific tenant database."""
dsn = TENANT_DB_URL_TEMPLATE.format(db_name=db_name)
try:
conn = psycopg2.connect(dsn, connect_timeout=timeout)
cur = conn.cursor()
cur.execute("""
SELECT
(SELECT COUNT(*) FROM employees WHERE is_active = true) as employees,
(SELECT COUNT(*) FROM inventory WHERE is_active = true) as inventory,
(SELECT COUNT(*) FROM customers WHERE is_active = true) as customers,
(SELECT COUNT(*) FROM sales WHERE created_at > NOW() - INTERVAL '30 days') as sales_30d,
pg_database_size(current_database()) as db_size
""")
row = cur.fetchone()
cur.close()
conn.close()
return {
"status": "ok",
"employees": row[0],
"inventory": row[1],
"customers": row[2],
"sales_30d": row[3],
"db_size_mb": round(row[4] / (1024 * 1024), 2)
}
except Exception as e:
return {"status": "error", "error": str(e)}

View File

@@ -0,0 +1,100 @@
"""Migration orchestration service."""
import os
import sys
POS_DIR = os.environ.get("POS_DIR", "/home/Autopartes/pos")
if POS_DIR not in sys.path:
sys.path.insert(0, POS_DIR)
from tenant_db import get_master_conn
from config import MIGRATIONS_DIR
def list_available_migrations():
"""List migrations found in POS migrations directory."""
migrations = []
if os.path.isdir(MIGRATIONS_DIR):
for fname in sorted(os.listdir(MIGRATIONS_DIR)):
if fname.endswith(".sql") and fname.startswith("v"):
version = fname.replace(".sql", "")
migrations.append({"version": version, "file": fname})
return migrations
def get_tenant_versions():
"""Get schema version for every tenant."""
conn = get_master_conn()
cur = conn.cursor()
cur.execute("""
SELECT t.id, t.name, t.db_name, COALESCE(v.version, 'v0.0') as version
FROM tenants t
LEFT JOIN tenant_schema_version v ON v.tenant_id = t.id
WHERE t.is_active = true
ORDER BY t.id
""")
results = []
for row in cur.fetchall():
results.append({
"tenant_id": row[0], "name": row[1], "db_name": row[2], "version": row[3]
})
cur.close()
conn.close()
return results
def run_migration_on_tenant(db_name, version):
"""Apply a single migration file to a tenant DB."""
from migrations.runner import apply_migration
return apply_migration(db_name, version)
def run_all_pending_migrations():
"""Run all pending migrations on all active tenants (wrapper around POS runner)."""
from migrations.runner import run_migrations
import io
import contextlib
# Capture stdout to return as log
f = io.StringIO()
with contextlib.redirect_stdout(f):
run_migrations()
return {"log": f.getvalue()}
def run_migration_on_all_tenants(version):
"""Apply one specific migration version to all tenants that don't have it."""
from migrations.runner import MIGRATIONS, apply_migration
conn = get_master_conn()
cur = conn.cursor()
cur.execute("""
SELECT t.id, t.db_name, COALESCE(v.version, 'v0.0') as version
FROM tenants t
LEFT JOIN tenant_schema_version v ON v.tenant_id = t.id
WHERE t.is_active = true
""")
tenants = cur.fetchall()
cur.close()
conn.close()
results = []
for tenant_id, db_name, current_version in tenants:
if current_version >= version:
results.append({"tenant_id": tenant_id, "db_name": db_name, "skipped": True, "reason": "already at or past version"})
continue
success = apply_migration(db_name, version)
if success:
# Update version tracker
conn2 = get_master_conn()
cur2 = conn2.cursor()
cur2.execute("""
INSERT INTO tenant_schema_version (tenant_id, version)
VALUES (%s, %s)
ON CONFLICT (tenant_id) DO UPDATE SET version = %s, updated_at = NOW()
""", (tenant_id, version, version))
conn2.commit()
cur2.close()
conn2.close()
results.append({"tenant_id": tenant_id, "db_name": db_name, "success": success})
return results

View File

@@ -0,0 +1,305 @@
"""Tenant management service wrapping POS tenant_manager."""
import os
import sys
import psycopg2
from psycopg2 import sql
from psycopg2.extensions import ISOLATION_LEVEL_AUTOCOMMIT
# Add POS to path so we can reuse tenant_manager
POS_DIR = os.environ.get("POS_DIR", "/home/Autopartes/pos")
if POS_DIR not in sys.path:
sys.path.insert(0, POS_DIR)
from config import MASTER_DB_URL, TENANT_DB_URL_TEMPLATE, DEMO_DEFAULT_DAYS
def get_master_conn():
return psycopg2.connect(MASTER_DB_URL)
def list_tenants(include_stats=False):
"""List all tenants with optional per-tenant stats."""
conn = get_master_conn()
cur = conn.cursor()
cur.execute("""
SELECT t.id, t.name, t.db_name, t.subdomain, t.rfc, t.plan, t.is_active,
t.created_at, COALESCE(s.expires_at, NULL) as expires_at,
COALESCE(v.version, 'v0.0') as schema_version
FROM tenants t
LEFT JOIN subscriptions s ON s.tenant_id = t.id
LEFT JOIN tenant_schema_version v ON v.tenant_id = t.id
ORDER BY t.id DESC
""")
cols = [desc[0] for desc in cur.description]
tenants = []
for row in cur.fetchall():
tenant = dict(zip(cols, row))
tenant["created_at"] = str(tenant["created_at"]) if tenant["created_at"] else None
tenant["expires_at"] = str(tenant["expires_at"]) if tenant["expires_at"] else None
tenant["is_demo"] = tenant["plan"] in ("demo", "trial")
tenant["demo_days_left"] = None
if tenant["expires_at"]:
from datetime import datetime
try:
exp = datetime.fromisoformat(tenant["expires_at"].replace("Z", "+00:00"))
now = datetime.now(exp.tzinfo) if exp.tzinfo else datetime.now()
tenant["demo_days_left"] = max(0, (exp - now).days)
except Exception:
pass
tenants.append(tenant)
cur.close()
conn.close()
if include_stats:
for t in tenants:
t["stats"] = _get_tenant_quick_stats(t["db_name"])
return tenants
def get_tenant(tenant_id):
"""Get single tenant details."""
conn = get_master_conn()
cur = conn.cursor()
cur.execute("""
SELECT t.id, t.name, t.db_name, t.subdomain, t.rfc, t.plan, t.is_active,
t.created_at, COALESCE(s.expires_at, NULL) as expires_at,
COALESCE(s.status, 'unknown') as subscription_status,
COALESCE(v.version, 'v0.0') as schema_version
FROM tenants t
LEFT JOIN subscriptions s ON s.tenant_id = t.id
LEFT JOIN tenant_schema_version v ON v.tenant_id = t.id
WHERE t.id = %s
""", (tenant_id,))
row = cur.fetchone()
cur.close()
conn.close()
if not row:
return None
keys = ["id", "name", "db_name", "subdomain", "rfc", "plan", "is_active",
"created_at", "expires_at", "subscription_status", "schema_version"]
return {k: str(v) if v is not None else None for k, v in zip(keys, row)}
def _get_tenant_quick_stats(db_name):
"""Quick stats for a tenant DB."""
dsn = TENANT_DB_URL_TEMPLATE.format(db_name=db_name)
try:
conn = psycopg2.connect(dsn, connect_timeout=5)
cur = conn.cursor()
cur.execute("""
SELECT
(SELECT COUNT(*) FROM employees WHERE is_active = true),
(SELECT COUNT(*) FROM inventory WHERE is_active = true),
(SELECT COUNT(*) FROM customers WHERE is_active = true),
(SELECT COUNT(*) FROM sales WHERE status = 'completed'),
pg_database_size(current_database())
""")
emp, inv, cust, sales, size = cur.fetchone()
cur.close()
conn.close()
return {
"employees": emp,
"inventory_items": inv,
"customers": cust,
"completed_sales": sales,
"db_size_mb": round(size / (1024 * 1024), 2)
}
except Exception as e:
return {"error": str(e)}
def create_demo(name, email, demo_days=None, subdomain=None, pin="0000"):
"""Provision a new demo tenant using POS tenant_manager."""
from services.tenant_manager import provision_tenant
from datetime import datetime, timedelta
days = demo_days or DEMO_DEFAULT_DAYS
if not subdomain:
from services.tenant_manager import generate_subdomain
subdomain = generate_subdomain(name)
# Ensure uniqueness by appending random suffix if needed
conn = get_master_conn()
cur = conn.cursor()
cur.execute("SELECT 1 FROM tenants WHERE subdomain = %s", (subdomain,))
if cur.fetchone():
import secrets
subdomain = f"{subdomain}-{secrets.token_hex(2)}"
cur.close()
conn.close()
result = provision_tenant(
name=name,
rfc=None,
owner_name="Admin Demo",
owner_email=email,
owner_pin=pin,
subdomain=subdomain
)
# Mark as demo plan and set expiration
tenant_id = result["tenant_id"]
conn = get_master_conn()
cur = conn.cursor()
cur.execute("UPDATE tenants SET plan = 'demo' WHERE id = %s", (tenant_id,))
cur.execute("""
INSERT INTO subscriptions (tenant_id, plan, status, expires_at)
VALUES (%s, 'demo', 'active', %s)
ON CONFLICT (tenant_id) DO UPDATE SET
plan = 'demo',
status = 'active',
expires_at = EXCLUDED.expires_at
""", (tenant_id, datetime.now() + timedelta(days=days)))
conn.commit()
cur.close()
conn.close()
result["demo_days"] = days
result["expires_at"] = str(datetime.now() + timedelta(days=days))
result["access_url"] = f"https://{subdomain}.nexusautoparts.com.mx/pos/login"
result["owner_pin"] = pin
return result
def reset_tenant(tenant_id, keep_config=True):
"""Reset a tenant: truncate business data but keep structure and owner."""
tenant = get_tenant(tenant_id)
if not tenant:
raise ValueError("Tenant not found")
db_name = tenant["db_name"]
dsn = TENANT_DB_URL_TEMPLATE.format(db_name=db_name)
tables_to_truncate = [
"inventory_operations",
"inventory",
"sale_items",
"sales",
"customer_payments",
"cash_register_closings",
"cash_register_movements",
"cash_registers",
"invoices",
"accounting_entries",
"journal_entries",
"service_orders",
"fleet_vehicles",
"crm_activities",
"quotations",
"quotation_items",
"savings_transactions",
"savings_accounts",
"supplier_orders",
"supplier_order_items",
"warranty_claims",
"notifications",
"inventory_uploads",
]
conn = psycopg2.connect(dsn)
cur = conn.cursor()
try:
for table in tables_to_truncate:
try:
cur.execute(f"TRUNCATE TABLE {table} RESTART IDENTITY CASCADE")
except Exception:
pass # Table may not exist
conn.commit()
success = True
except Exception as e:
conn.rollback()
success = False
raise RuntimeError(f"Reset failed: {e}")
finally:
cur.close()
conn.close()
return {"success": success, "tenant_id": tenant_id, "tables_reset": len(tables_to_truncate)}
def delete_tenant(tenant_id):
"""Permanently delete a tenant and its database."""
tenant = get_tenant(tenant_id)
if not tenant:
raise ValueError("Tenant not found")
db_name = tenant["db_name"]
conn = get_master_conn()
cur = conn.cursor()
# Drop database
try:
master_conn = psycopg2.connect(MASTER_DB_URL)
master_conn.set_isolation_level(ISOLATION_LEVEL_AUTOCOMMIT)
master_cur = master_conn.cursor()
master_cur.execute(
sql.SQL('DROP DATABASE IF EXISTS {}').format(sql.Identifier(db_name))
)
master_cur.close()
master_conn.close()
except Exception as e:
pass
# Clean master records
cur.execute("DELETE FROM tenant_schema_version WHERE tenant_id = %s", (tenant_id,))
cur.execute("DELETE FROM subscriptions WHERE tenant_id = %s", (tenant_id,))
cur.execute("DELETE FROM tenants WHERE id = %s", (tenant_id,))
conn.commit()
cur.close()
conn.close()
return {"success": True, "tenant_id": tenant_id, "db_name": db_name}
def toggle_tenant(tenant_id, active):
"""Activate or deactivate a tenant."""
conn = get_master_conn()
cur = conn.cursor()
cur.execute("UPDATE tenants SET is_active = %s WHERE id = %s", (active, tenant_id))
conn.commit()
rowcount = cur.rowcount
cur.close()
conn.close()
return {"success": rowcount > 0, "tenant_id": tenant_id, "is_active": active}
def get_tenant_login_url(subdomain):
"""Generate login URL for a tenant."""
domain = os.environ.get("NEXUS_DOMAIN", "nexusautoparts.com.mx")
return f"https://{subdomain}.{domain}/pos/login"
def get_dashboard_stats():
"""Global stats for the manager dashboard."""
conn = get_master_conn()
cur = conn.cursor()
cur.execute("SELECT COUNT(*) FROM tenants")
total = cur.fetchone()[0]
cur.execute("SELECT COUNT(*) FROM tenants WHERE is_active = true")
active = cur.fetchone()[0]
cur.execute("SELECT COUNT(*) FROM tenants WHERE plan = 'demo'")
demos = cur.fetchone()[0]
cur.execute("""
SELECT COUNT(*) FROM subscriptions
WHERE status = 'active' AND expires_at < NOW() + INTERVAL '7 days'
""")
expiring_soon = cur.fetchone()[0]
cur.close()
conn.close()
# Get system health summary
from services.health_service import check_disk_space, check_memory
disk = check_disk_space()
mem = check_memory()
return {
"tenants": {"total": total, "active": active, "demos": demos, "expiring_soon": expiring_soon},
"system": {
"disk_percent": disk.get("percent_used"),
"memory_percent": mem.get("percent_used"),
"disk_free_gb": disk.get("free_gb"),
"memory_available_gb": mem.get("available_gb")
}
}