问题
I am trying to develop a small app which will gather weather data from an API. I have used APScheduler to execute the function every x minutes. I use Python Tornado framework.
The error I am getting is:
INFO Job "GetWeather (trigger: interval[0:01:00], next run at: 2015-03-28 11:40:58 CET)" executed successfully
ERROR Exception in callback functools.partial(<function wrap.<locals>.null_wrapper at 0x0335C978>, <tornado.concurrent.Future object at 0x03374430>)
Traceback (most recent call last):
File "C:\Python34\Lib\site-packages\tornado\ioloop.py", line 568, in _run_callback
ret = callback()
File "C:\Python34\Lib\site-packages\tornado\stack_context.py", line 275, in null_wrapper
return fn(*args, **kwargs)
greenlet.error: cannot switch to a different thread
Which I think is coming from the Coroutine from GetWeather() as, if I remove all asycn features from it, it works.
I am using Motor to read needed coordinates and pass them through the API and store weather data in MongoDB.
import os.path, logging
import tornado.web
import tornado.ioloop
from tornado.httpclient import AsyncHTTPClient
from tornado import gen
from tornado.options import define, options
from apscheduler.schedulers.tornado import TornadoScheduler
import motor
client = motor.MotorClient()
db = client['apitest']
console_log = logging.getLogger(__name__)
define("port", default=8888, help="run on the given port", type=int)
define("debug", default=False, help="run in debug mode")
class MainRequest (tornado.web.RequestHandler):
def get(self):
self.write("Hello")
scheduler = TornadoScheduler()
class ScheduledTasks(object):
def get(self):
print("This is the scheduler");
def AddJobs():
scheduler.add_job(GetWeather, 'interval', minutes=1)
def StartScheduler():
scheduler.start();
def StopScheduler():
scheduler.stop();
class Weather(tornado.web.RequestHandler):
def get(self):
self.write("This is the Weather Robot!")
GetWeather()
@gen.coroutine
def GetWeather():
'''
Getting city weather from forecast.io API
'''
console_log.debug('Start: weather robot')
cursor = FindCities()
while (yield cursor.fetch_next):
city = cursor.next_object()
lat = str(city["lat"])
lon = str(city["lon"])
http_client = AsyncHTTPClient()
response = yield http_client.fetch("https://api.forecast.io/forecast/3925d0668cf520768ca855951f1097cd/%s,%s" %(lat, lon))
if response.error:
print ("Error:", response.error)
# Store all cities with errors in order to save them in the log file
else:
json = tornado.escape.json_decode(response.body)
temperature = json["currently"]["temperature"]
summary = json["currently"]["summary"]
db.cities.update({'_id': city["_id"]}, {'$set': {'temperature': temperature, 'summary': summary}})
console_log.debug('End: weather robot')
return
def FindCities():
'''
cities = [{
"_id" : ObjectId("55165d07258058ee8dca2172"),
"name" : "London",
"country" : "United Kingdom",
"lat" : 51.507351,
"lon" : -0.127758
},
{
"_id" : ObjectId("55165d07258058ee8dca2173"),
"name" : "Barcelona",
"country" : "Spain",
"lat" : 41.385064,
"lon" : 2.173403
}
'''
cities = db.cities.find().sort([('_id', -1)])
return cities
def main():
logging.basicConfig(level=logging.DEBUG,format='%(levelname)-8s %(message)s')
app = tornado.web.Application(
[
(r'/robots/weather', Weather),
(r'/', MainRequest)
],
cookie_secret="__TODO:_GENERATE_YOUR_OWN_RANDOM_VALUE_HERE__",
login_url="/auth/login",
template_path=os.path.join(os.path.dirname(__file__), "templates"),
static_path=os.path.join(os.path.dirname(__file__), "static"),
xsrf_cookies=True,
debug=options.debug,
)
app.listen(options.port)
AddJobs()
StartScheduler()
tornado.ioloop.IOLoop.instance().start()
if __name__ == "__main__":
main()
Any idea what I am doing wrong? As I see in APScheduler code, TornadoScheduler() runs in Tornado IOLoop... (https://bitbucket.org/agronholm/apscheduler/src/a34075b0037dba46735bae67f598ec6133003ef1/apscheduler/schedulers/tornado.py?at=master)
Oh! I forgot to say that the idea is to be able to execute the task via APScheduler or manually both.
Many thanks!
回答1:
By default, TornadoScheduler runs scheduled tasks in a thread pool. Your specific task, however, uses the IOLoop and so expects to be run in the same thread. To fix this, you can use the add_callback() method of the tornado IOLoop to schedule a task to be run in the IOLoop's thread as soon as possible.
Like so:
def your_scheduled_task():
IOLoop.instance().add_callback(your_real_task_function)
or even better:
scheduler.add_job(IOLoop.instance().add_callback, 'interval', minutes=1, args=[GetWeather])
回答2:
Looks to me like TornadoScheduler, even though it integrates with the IOLoop, it still runs operations on a thread pool:
def _create_default_executor(self):
"""Creates a default executor store, specific to the particular scheduler type."""
return ThreadPoolExecutor()
Motor doesn't like running on multiple threads in the same process--I only test use cases where Motor is used on the main thread.
I think you should either use a Tornado PeriodicCallback instead of APScheduler, or you should use PyMongo with APScheduler (so PyMongo is running on background threads) instead of Motor.
来源:https://stackoverflow.com/questions/29316173/apscheduler-run-async-function-in-tornado-python