Hello,

I'm relatively new to nginx so I may be completely wrong about what I
assume here. Please excuse me if it is the case.

I'm streaming large quantities of data from a FCGI application through
nginx (large = several gigabytes). When streaming the data, nginx
starts to use more and more memory, with the worker process taking
more than a gigabyte of RAM at a point. The more I stream, the more
the nginx process use. I guess it should be categorized as a bug for
several reasons :

- It should not happen. nginx should not use that much memory when
streaming data from a FCGI application.
- It may even lead to DoS in a shared hosting situation. Someone on
the server may create a FCGI application which streams a lot of data,
making the nginx process take a lot of memory and finally be killed by
the kernel (OOM). I guess the worker would be restarted by the master
process in these cases, but it would exhaust memory and probably swap,
which is never good.

A simple python+flup script reproducing this problem :

def data():
while True:
yield ('x' * 4096)

def myapp(environ, start_response):
start_response('200 OK', [('Content-Type', 'text/plain')])
return data()

if __name__ == '__main__':
from flup.server.fcgi import WSGIServer
WSGIServer(myapp, bindAddress=('127.0.0.1',31337)).run()

When requesting the data from localhost (curl http://localhost:port/
>/dev/null) I got a memory usage increasing by more than 10M/s, which
really is a lot for something that should not in the first place
exhaust memory.

As I said, I'm new to nginx and this may be wanted "by design". I
personally think this is a bug.

Regards,

--
Pierre "delroth" Bourdon <delroth@gmail.com>
Étudiant à l'EPITA / Student at EPITA
http://bitbucket.org/delroth/

_______________________________________________
nginx-devel mailing list
nginx-devel@nginx.org
http://nginx.org/mailman/listinfo/nginx-devel