[Bese-devel] : ANN: UCW-CORE is ready. The future is now!

[Drew Crampsie]

2009/1/4 Aycan iRiCAN <aycan.irican at core.gen.tr>:
> Mr. Crapsie,

"Cra_m_psie" ... let's not make it worse than it is. :)

> Could you please give us more info about your production environment?

You bet. The application is not terribly busy (~5 simul. users), But
does cons a lot (searches and reports), keeps a lot in RAM.

I am currently running a single threaded ucw httpd behind mod_proxy.
SBCL, linux x86 Xen 32 bit 512mb RAM (on the tech co-op's Xen
platform). The excutable is created without ldb. this is important, as
it is still blowing the heap about once a week... though the users
have not  really noticed ;)

I run the process under runit, and init-like program with a service
monitor that will restart the process if  (when) it crashes. The
crashes are not due to any fault of ucw per-se, but rather the fact
that i'm keeping a lot of search results around in components, and
sometimes the users return a lot of results. It's.usually a GC that
kills it. I've identified the worst offenders and made some changes,
but i've yet to deploy them under real world situations.

I will likely deploy behind mod_lisp when we do a point release. Also.
running multiple processes behind a proxy with 'sticky sessions'  is
an option if the application gets busier before i deal with the memory
hogs :).

The software itself tracks payments on insurance claims, is built on
postgresql, rofl (my own o/r mapping) and lisp on lines. It is a
version 2.0 of the very first UCW application i wrote, and has
travelled from ucw 0.3 or so (remember when we did releases?) to
ucw_dev, through ucw_ajax and now runs on ucw-core.

Cheers,

drewc


>
> Best Regards,
> --
> aycan at core
>