[Zope-dev] Critical Bug?

[Lars Hoss]

hi all!

today we found a critical bug in zope and we would like
to know wether anyone knows a solution.

it seems that zope crashes under heavy load (which means
more than 15 hits/sec) and beeing authenticated against
the zope server.

try for example the following:

authenticate against your zope server, e.g http://some.server.com/manage.
afterwards view a page on this server with some dtml code.
now reload this page very fast and very often by pressing F5 permanently.
after a little while (between 30 seconds and 2 minutes) the connect
may fail. on the server side we saw for example the following message:

2000-02-11T16:07:06 ERROR(200) ZServer uncaptured python exception, closing
channel <zhttp_server listening :8080 at 860d2d8>
(exceptions.TypeError:unpack non-sequence
[/opt/shares/volume2/zope/Zope-2.1.4-src/ZServer/medusa/asyncore.py|poll|59]
[/opt/shares/volume2/zope/Zope-2.1.4-src/ZServer/medusa/asyncore.py|handle_r
ead_event|285]
[/opt/shares/volume2/zope/Zope-2.1.4-src/ZServer/medusa/http_server.py|handl
e_accept|609])

we tried to connect from another maschine and successed. but the maschine
who
failed before is in most cases not able to reconnect for a longer time.

we were able to reproduce this behavior many times with IE 5.0.

our python developer added the code self.close() to
ZServer/medusa/asyncore.py at line 320
where self.close() is:

def close (self):
     self.del_channel()
     self.socket.close()

afterwards the zserver seems to work stable. we have to test out a bit more.

cheers,
lars