[uWSGI] how to properly set nodes and udp in cache2 config

Roberto De Ioris roberto at unbit.it
Thu Oct 17 17:37:39 CEST 2013


> yes, the master and slave are two different servers. the updates can
> be propagated to slave after I use the following settings:
>
> on master (ip=10.1.1.11):
> cache2 =
> name=test-conf-cache,items=8192,blocksize=2048,store=/shared_disk/test_cache2.dump,nodes=10.1.1.12:7172,udp=10.1.1.11:7171
>
> and on slave (ip=10.1.1.12):
> cache2 =
> name=test-conf-cache,items=8192,blocksize=2048,store=/shared_disk/test_cache2.dump,udp=10.1.1.12:7172
>
> now everything seems alright. but i am wondering if there will be some
> problem in the future since you mentioned that i should not use the
> store=/shared_disk/test_cache2.dump as a part of the config on slave
> node. I tried setting sync, but failed. In fact, i cannot make the
> master as a cache2 sync server because it works behind nginx via unix
> sockets. The slave node cannot connect to the sync server whatever
> sync address i set. So i guess to restore the initial data from a file
> on shared disk is my only choice. Is this setting risky? what do you
> think?
>
> thanks,
>

The problem is that if you sync while an update is in place you could get 
an inconsistent cache image.

I would try understanding why sync=<node> does not work for the slave.

Remember that sync happens via tcp so the <node> address is the same value
of --socket of the uwsgi instance
-- 
Roberto De Ioris
http://unbit.it


More information about the uWSGI mailing list