Airflow Benchmark

1. Test Environment

Hostmachine:

System	Linux Kernel Version	CPU	Ram	Airflow version
Kubuntu 20.04	5.4.0-67-generic	Intel 7700HQ 4c8t	16GB	2.01

Network：

Network	Bandwidth	Router
CERNNET East China	1000Mbps	Redmi AC2100

2. Concurrent test

Condition: using python script set a 10000 thread pool to send POST request to the airflow server.

import requests
import json
import threading
from pprint import pprint
import random
webserver_user = ""
webserver_passwd = ""
MAX_CONCURRENT_NUM = 500
SERVER_ADDRESS = "http://localhost:8080/api/v1/"
DAG_DIR = "dags/example_external_task_marker_child/dagRuns"


def make_a_request():
    random_ID = str(random.randint(7, 19991))
    base_json = {"dag_run_id": random_ID}
    result = requests.post(SERVER_ADDRESS+DAG_DIR
        ,
        data=json.dumps(base_json),
        auth=(webserver_user,webserver_passwd), headers={'Content-Type': 'application/json'})
    pprint(result.content.decode('utf-8'))


threadpool = []
for i in range(MAX_CONCURRENT_NUM):
    th = threading.Thread(target=make_a_request, args=())
    threadpool.append(th)
for th in threadpool:
    th.start()
for th in threadpool:
    threading.Thread.join(th)

The Airflow Scheduler output

DAG example_external_task_marker_child already has 29 active runs, not queuing any tasks for run 2021-03-22 08:08:27.212848+00:00

It means that the max size of the task queue is 29. Maybe could be modified to support more requests.

The Max_request_number

The maximum_page_limit in the official doc is 100. In the reality test, when the MAX_CUCURRENCY_NUMBER up to 2000, I got a request error.