pylibmc is not thread-safe

[cedk]

The way pylibmc client is used is not thread-safe.
I think flask-caching should use the ThreadMappedPool.
Otherwise on threaded Flask application, we must use python-memcached but there is no way to force to use this library.

[reubano]

I think I've run into this issue. While running on heroku, not all instances of the cache respect CACHE_DEFAULT_TIMEOUT. It took me ages to figure out why my cache was clearing after 5 minutes, despite having set CACHE_DEFAULT_TIMEOUT for longer.

[reubano]

I'm using python-binary-memcached>=0.26.1,<0.27.0

[sh4nks]

Actually this is not a bug. If you need a threadsafe memcache backend, you can just subclass the existing one and add it there. The reason I am not adding it is as due to the fact that the ThreadMappedPool is only supported by pylibmc and we are supporting 4 different memcache libraries.

[cedk]

Then maybe pylibmc should not be the first library to use as its client is not thread safe by default.

[sh4nks]

Yeah maybe. I am happy to review a PR if you want to submit one :-)

[reubano]

@cedk can you outline the steps you took to get thread safe behavior with python-memcached?

[sh4nks]

I am closing this issue for now. As I have said before, I'd happily review a PR if someone wants to implement this.