[OpenAFS] Re: OpenAFS-devel] Windows client - memory leak?

[Michael Lasevich]

Not really, I am still getting all sorts of grief.

A partial workaround to a lot of issues was to have Bill Richardson (of
RHIT) include some special code to restart AFS service each time the token
is renewed in his WAKE application
(http://www.rose-hulman.edu/TSC/software/wake/ - which I think anyone
running AFS for windows should have, especially if they want to integrate
authentication between Win2k and AFS) It's a somewhat hidden option in the
registry, but it really saves a lot of grief.

Generally speaking I find Windows openAFS 1.2.2b somewhat lacking the
stability  to be used in the production environment (at least for running on
Win2K) . Every  machine I loaded it onto almost immediately started
exibiting erratic behaviour. In many cases I had to switch AFS service to
manual startup and let WAKE start it up on load. This stopped most of the
boot-time problems -- in "Automatic" startup mode, 1.2.2b had about 1 in 3
failure rate on statartup. But that still leaves some problems. Occasionally
AFS will not get restarted by WAKE , at that point the only way to get AFS
working is reboot. But the biggest issue is that when AFS is running it
still dows something strange to memory, because I have all sorts of
applications crash with the message like this (comming from different
applications):

MSTask.exe - Application Error : The instruction at "0x77fcb03d" referenced
memory at "0x00000000". The memory could not be "written".

This seems to me that Which seems to me like maybe some buffer overflow
problems exist in the client which would account for memory leakage and
possibly other application crahes (if windows indeed does allow one
application to get into memory space of another, which would not surprise me
at all)

All in all, given these issues and given that I do not believe anyone is
really supporting or cares for Windows client(most of the questions about
Windows client to this list seem to go unanswered, at least publically)  I
decided to hold off on deploying windows part of the AFS here indefinately.
At some point I am going to check cout the latest code from CVS and play
with it, but right now I simply do not have the time for that. And to add to
that id does not appear that anyone really maintaines or cares much for the
windows clients.

I have to add some disclaimers to this, all of the tests I've done were done
on identically configured machines, with pretty much same software set
including DB2 client and DB2Connect, which some people on this list though
DB2 stuff may cause problems
for openAFS (not sure why and it was never confirmed or denied) Also, I've
only tried using openAFS client on windows, not server.


-michael


----- Original Message -----
From: "Christian Boissat" <Christian.Boissat@cern.ch>
To: <openafslist@lasevich.net>
Sent: Monday, July 22, 2002 2:52 AM
Subject: OpenAFS-devel] Windows client - memory leak?


> Michael and All,
>
> At CERN, for some openAFS end-users, we have the same behaviour as the one
> described in :
>
>     https://lists.openafs.org/pipermail/openafs-devel/2002-May/002936.html
>
> To your knowledge, has the problem being identified. Is there any fix in
> preparation ?
>
> Cheers; Christian Boissat
>
>