/docs/processes.txt

Processes
---------

.. warning::

   I'm pretty sure my understanding of the process model is actually
   incorrect, and this document isn't correct.

This document explains the process model of Silver Lining.

Silver Lining runs under `mod_wsgi
<http://code.google.com/p/modwsgi/>`_, but mod_wsgi itself is
configurable, and Silver Lining configures it a specific way.  The
result works like this:

* Each application instance runs in its own processes.  So if you
  upload multiple applications (that are all live simultaneously) then
  those will never be in the same process.  This is necessary because
  there's no way besides separate processes to keep paths and
  libraries straight.

* An application gets 5 concurrent processes.  That is, there is a
  pool of 5 processes that wait for a request to come in.

* There are no threads, each process is single-threaded.

* Processes live for a limited number of requests (200 requests).
  After that the process is killed and another started in its place.
  This avoids some possible memory leaks or other problems associated
  with long-running processes.

* All processes run as the user ``www-data``.

* "Internal" requests are run in their own short-lived processes.
  This includes requests like the ``update_fetch`` request.  They
  otherwise run very similarly to normal requests (as ``www-data``,
  similar environment, etc).  However, because it is not actually
  under mod_wsgi, it is very possible that there will be differences
  in the environ.