On Fri, Oct 24, 2014 at 3:49 PM, Mikko Ohtamaa <[email protected]>
wrote:

>
>
>
>> Some background: For most people a worker going down isn't the end of the
>> world, but our app uses threads which acquire locks through Memcached (we
>> use a load balancer and multiple app servers), so if a request gets killed
>> before it can release the lock, that lock stays held forever; e.g. our
>> (albeit experimental) app has been out of commission since Tuesday.
>> Incidentally, I'm all ears if you have suggestions on handling that more
>> resiliently :)
>>
>
> Try having timeouting locks. E.g.
> https://pypi.python.org/pypi/DistributedLock/ This mitigates the issue
> somehow.
>

Are you using that project, and does it work well for you? I looked at it,
but then I found out redis-py has a distributed lock that seems more
robust. We might decide to switch to Redis just for that.

Thanks,
Matt
_______________________________________________
uWSGI mailing list
[email protected]
http://lists.unbit.it/cgi-bin/mailman/listinfo/uwsgi

Reply via email to