← Back to team overview

openstack team mailing list archive

Re: Architecture for Shared Components

 

On Aug 4, 2010, at 10:17 AM, Michael Gundlach wrote:

On Wed, Aug 4, 2010 at 9:29 AM, Jorge Williams <jorge.williams@xxxxxxxxxxxxx<mailto:jorge.williams@xxxxxxxxxxxxx>> wrote:
Can you configure WSGI  to split things up as needed for a particular deployment(?)

WSGI itself doesn't provide that (it just defines an interface for dispatching/receiving requests) but we can build that if it hasn't already been built by someone solving the same problem (which I'd bet dollars to donuts someone has.)

K. Cool.


Perhaps a deployment configuration file could specify which layers go in which order and which ones get stacked in a single server.  Then to start a server the command would be something like

proxyserver --wsgi-layers-top-down=some.wsgi.app,another.wsgi.app,yet.another.wsgi.app


I used to work on a project back-in-the-day that had very similar requirements.

Configuration was itself written in a python script:


proxy1 = Proxy1()
proxy2 = Proxy2()

proxy1.Conf(.....)
proxy2.Conf(.....)

app = App()

app.AddNode(proxy1)
app.AddNode(proxy2)

app.go()

You executed the configurations script to start the application. The app came bundled with a whole bunch of scripts that people could customize.  There was a network protocol involved  in transmitting configuration options to other nodes. Will this work?

FYI:  The project I'm referring to is the Chromium project (http://chromium.sourceforge.net/doc/index.html).  It's a cluster based rendering system.  See the section on configuration scripts.   The app was written in C, but all configuration was done in python as described above. Eventually somebody developed a graphic tool where you can point and click things together and it would spit out python configuration code :-)

-jOrGe W.


Follow ups

References