pyapache anyone

[Sam Penrose]

Nell'articolo <mailman.988225957.759.python-list at python.org>, "Sam
Penrose" <sam at ddmweb.com> ha scritto:

> PyApache 4.19 was broken with a memory leak that appeared after on the
> order of 100,000 dynamic requests on our site (RedHat 6.1 on Intel,
> Apache1.3.9 Red-Hat-Secure/3.1 mod_ssl/2.4.10 OpenSSL/0.9.4) . I sent
> Lele a bug report at the address listed in the docs a year ago; the
> address he gave at that time (lele at integra.it) failed with a "too many
> hops" message. There was also a mailing list that didn't work. I finally
> posted to comp.lang.python and got confirmation from another user.
> 
> Lele, did you ever get wind of such a leak?
> 

Lele ha scritto:
> No, I wasn't aware of it. Do you have any simple/self-contained script
> that triggers the problem?

Nope. We used it for months in preproduction, then went live and the
site was a lot busier than anyone expected. The failure was
catastrophic, and all we were interested in was making the site work
again. We've since gone to a cluster.

Coincidentally, google just restored Deja's Usenet archive, so you can
see the thread I spawned and the confirmation I got at the end of it
from "seant at iname.com":

"I've just completed a couple of overnight thrashing tests. Serving
400,000 pages or so appears to leak about 400MB of memory.
This amounts to about 1kb per page invocation. Each page was served
by PyApache. The system tracking tool on Linux showed the 10 or so
httpd processes that had been spawned as having grown to about 40MB
each."

<http://groups.google.com/groups?hl=en&lr=&safe=off&ic=1&th=5f604e35e318d1e4&seekm=8dni3s%241o1%241%40nnrp1.deja.com#p>

(or search for PyApache in comp.lang.python in April of last year).