feat: DAG backfill Metrocuadrado CO

This commit is contained in:
2026-08-02 18:05:32 +02:00
parent 30b4b33ec7
commit b6648d3234
+57
View File
@@ -0,0 +1,57 @@
"""DAG de backfill Metrocuadrado CO (Colombia)
Reanudable via checkpoint en /opt/scrapers/metrocuadrado_co/state.json
156 categorías × (1 pagina + ~24 barrios) × ~55 listings cada una.
"""
from datetime import datetime, timedelta
import paramiko
from airflow import DAG
from airflow.operators.python import PythonOperator
VPS_HOST = '152.53.242.31'
VPS_USER = 'root'
VPS_KEY = '/opt/airflow/config/.ssh/id_rsa'
SCRAPER_CMD = (
'timeout 36000 '
'/home/jpuma/proyectos/bot-sunat/.venv/bin/python '
'/opt/scrapers/metrocuadrado_co/scraper.py 2>&1'
)
def run_backfill(**ctx):
client = paramiko.SSHClient()
client.set_missing_host_key_policy(paramiko.AutoAddPolicy())
client.connect(VPS_HOST, username=VPS_USER, key_filename=VPS_KEY, timeout=30)
_, stdout, stderr = client.exec_command(SCRAPER_CMD, timeout=36000)
output = stdout.read().decode('utf-8', errors='replace')
err = stderr.read().decode('utf-8', errors='replace')
client.close()
print(output[-6000:] if len(output) > 6000 else output)
if err:
print('STDERR:', err[-500:])
if 'Finalizado:' not in output:
raise RuntimeError('Scraper no imprimió "Finalizado:" — posible error')
return output.split('Finalizado:')[-1].strip()
default_args = {
'owner': 'jpuma',
'retries': 1,
'retry_delay': timedelta(minutes=15),
'execution_timeout': timedelta(hours=11),
}
with DAG(
dag_id='metrocuadrado_co_backfill',
default_args=default_args,
description='Backfill histórico Metrocuadrado CO — reanudable via checkpoint',
schedule_interval=None,
start_date=datetime(2026, 8, 1),
catchup=False,
max_active_runs=1,
tags=['colombia', 'backfill', 'inmuebles'],
) as dag:
PythonOperator(
task_id='backfill_metrocuadrado_co',
python_callable=run_backfill,
)