I'm using:
uWSGI start params:
--socket 127.0.0.1:8081 --daemonize --enable-threads --threads 2 --processes 2
I setup my MongoClient ONE time:
self.mongo_client = MongoClient('mongodb://user:[email protected]:port/mydb')
self.db = self.mongo_client['mydb']
I try and save a JSON dict to MongoDB:
result = self.db.jobs.insert_one(job_dict)
It works via a unit test that executes the same code path to mongodb. However when I execute via CherryPy and uWSGI using an HTTP POST, I get this:
pymongo.errors.ServerSelectionTimeoutError: No servers found yet
Why am I seeing this behavior when run via CherryPy and uWSGI? Is this perhaps the new thread model in PyMongo 3?
Update:
If I run without uWSGI and nginx by using the CherryPy built-in server, the insert_one()
works.
Update 1/25 4:53pm EST:
After adding some debug in PyMongo, it appears that topology._update_servers()
knows that the server_type = 2 for server 'myserver-a.mongolab.com'. However server_description.known_servers()
has the server_type = 0 for server 'myserver.mongolab.com'
This leads to the following stack trace:
result = self.db.jobs.insert_one(job_dict)
File "/usr/local/lib/python3.4/site-packages/pymongo/collection.py", line 466, in insert_one
with self._socket_for_writes() as sock_info:
File "/usr/local/lib/python3.4/contextlib.py", line 59, in __enter__
return next(self.gen)
File "/usr/local/lib/python3.4/site-packages/pymongo/mongo_client.py", line 663, in _get_socket
server = self._get_topology().select_server(selector)
File "/usr/local/lib/python3.4/site-packages/pymongo/topology.py", line 121, in select_server
address))
File "/usr/local/lib/python3.4/site-packages/pymongo/topology.py", line 97, in select_servers
self._error_message(selector))
pymongo.errors.ServerSelectionTimeoutError: No servers found yet
We're investigating this problem, tracked in PYTHON-961. You may be able to work around the issue by passing connect=False when creating instances of MongoClient. That defers background connection until the first database operation is attempted, avoiding what I suspect is a race condition between spin up of MongoClient's monitor thread and multiprocess forking.