Thanks. Is it like the shared_buffer of PostgreSQL or SGA of Oracle (I am a database guy). The buffer is shared by all connections. So irrespective of the number of workers, this will be allocated only once? If I have 4 workers and set address space limit to 256 MB, all the workers together will not cross 256 MB?
On Wed, Nov 27, 2013 at 3:39 PM, Roberto De Ioris <[email protected]> wrote: > > > Hi, > > Digging into why my workers keep ding, I see that the address space is > > crossing the threshold. When does that happen? > > writev(2, [{"{address space usage: 277671936 bytes/264MB} {rss usage: > > 24555520 bytes/23MB} ", 78}, > > Regards, > > Jayadevan > > _______________________________________________ > > uWSGI mailing list > > [email protected] > > http://lists.unbit.it/cgi-bin/mailman/listinfo/uwsgi > > > > It is a pretty low-level concept: > > http://en.wikipedia.org/wiki/Virtual_memory > > you could see it as the "view" your process has of the memory (that does > not map 1:1 to physical memory) > > for example when you create a uWSGI cache, it is allocated in the master, > but shared between all of the workers. The master will account RSS for it > while workers will have it in the address space (i have over-simplified > but should clarify things) > > The same happens for shared libraries, mmapped devices and so on > > uWSGI allows you to limit it as it is (well, was) a common pattern for > ISPs for avoiding users to exhaust system resources (as limiting RSS is > not supported on the vast majority of OS) > > -- > Roberto De Ioris > http://unbit.it > _______________________________________________ > uWSGI mailing list > [email protected] > http://lists.unbit.it/cgi-bin/mailman/listinfo/uwsgi >
_______________________________________________ uWSGI mailing list [email protected] http://lists.unbit.it/cgi-bin/mailman/listinfo/uwsgi
