[racket] Running into severe scaling issues with plt-web-server

From: Jay McCarthy (jay.mccarthy at gmail.com)
Date: Wed Apr 10 18:42:13 EDT 2013

On Wed, Apr 10, 2013 at 4:21 PM, Danny Yoo <dyoo at hashcollision.org> wrote:
> Hi Jay,
>
>
> Followup: for now, it looks like we're just that we're being hammered
> hard enough that the web server has no choice but to starts sending
> 503 responses.  Memory usage appears bounded, despite what I had
> thought earlier.  I'll add additional server instances to host
> wescheme.org's compiler servers and follow up with observations.

Okay

> What is the current behavior of the system with regards to how it
> handles a barrage of requests?  I took a look at:
>
>     http://docs.racket-lang.org/web-server-internal/limit.html
>
> which describes what one might do to apply limits to the web server.
> However, I can not find documentation that states what the default
> rate-limiting behavior is when using serve/servlet.

The server makes no attempt to rate limit at all.

> Would you be amendable to adding a #:max-waiting keyword argument to
> serve/servlet to override the default backlog of 40 waiters?

Definitely. I won't be able to add it until tomorrow though afternoon
though... if you want to push before then.

>  The
> other calls into the web server provide a #:max-waiting parameter, and
> I'm not sure I understand why it's absent from serve/servlet.

It's absent because there's already 400 other keyword arguments :)

Jay


--
Jay McCarthy <jay at cs.byu.edu>
Assistant Professor / Brigham Young University
http://faculty.cs.byu.edu/~jay

"The glory of God is Intelligence" - D&C 93

Posted on the users mailing list.