[typo] Typo blog using 100MB RAM per process?

Jason Hoffman jason at jasonhoffman.org
Mon Oct 24 14:46:38 EDT 2005


On Oct 24, 2005, at 10:24 AM, Scott Laird wrote:
>
> Fun.  I'll see if I can reproduce any of this shortly.  I'd like to
> get this fixed today if at all possible.

I think the main issue is that rails dispatch.fcgi's aren't getting  
killed off when a lighttpd is restarted, and people in our forum have  
said the same. It's odd because we don't see it with any of our  
stuff. I don

So what happens is site something get's funny, then a daedalus  
doesn't pull a string on the site that it was checking for, so it  
restarts a lighttpd and it's fcgi without getting rid of the old  
ones, then the entire things doesn't really work and it then it just  
spirals up.

And in case anyone cares the Litespeed devs have a toggle to turn off  
persistent fcgi in their latest version of their enterprise server.  
George said,

"First, we found that the FCGI interface for RubyOnRails did not like
persistent FCGI connections (design issue), and has serious file
descriptor leaks and performance issues when persistent connection is
used. Both Apache and lighttpd does not take advantage of persistent
FCGI connections, so it was not issue when using RubyOnRails with them.
But personally, I think persistent connection should be supported to get
better performance. In order to work around this issue before the ruby
fcgi interface get fixed, an option has been added to turn off
persistent FCGI connection in litespeed. So, if by any chance, you want
to use LiteSpeed with RubyOnRails, remember to turn persistent
connection off."

- Jason


More information about the Typo-list mailing list