Quickie here that needs more domain expertise on pymongo than I have right now:
Are the "right" parts of the pymongo driver written in python for me to call gevent monkey_patch() and successfully alter pymongo's blocking behavior on r/w within gevent "asynchronous" greenlets?
If this will require a little more leg work on gevent and pymongo -- but it is feasible -- I would be more than willing to put in the time as long as i can get a little guidance over irc.
Thanks!
Note: At small scale mongo writes are not a big problem because we are just queuing a write "request" before unblocking. BUT talking to fiorix about his twisted async mongo driver (https://github.com/fiorix/mongo-async-python-driver), even mongo's quick write (requests) can cause problems in asyncronous applications at scale. (And of course, non-blocking reads could cause problems from the start!)
I have used PyMongo with Gevent and here are a few things you need to watch out for:
- Instantiate only one
pymongo.Connection
object, preferrably as a global or module-level variable. This is important becauseConnection
has within itself a pool! - Monkey patch everything, or at least BOTH socket and threading. Due to the use of thread locals in
Connection
, patching socket alone is not enough. - Remember to call
end_request
to return the connection to the pool.
The answer to your question is go ahead, PyMongo works just fine with Gevent.
On initial inspection it doesn't appear to do any socket operations in the c code so it should be fine (blocking ops should just block the green thread).
来源:https://stackoverflow.com/questions/7166998/pymongo-gevent-throw-me-a-banana-and-just-monkey-patch