[OpenAFS-devel] OpenAFS server 1.3.80 on x86_64

[Hartmut Reuter]

On opterons with SLES9 we had a problem that the epoch for the 
rx-connection was 0x80000000 which comes from an timeofday call which 
returned zero. This confused the fileservers which finds the connection 
using the epoch. With many of these opterons (180) he always ended up 
with the wrong connection with the effect that the clients got 
connection timed out.

But this seems to be a different problem. You may use rxdebug to verify 
that your client's epochs are correct.

Hartmut Retuer

Ulrich Schwickerath wrote:
> Hi, again,
> 
> sorry for the long period of silence from my side, I was mostly out of office 
> this week, and only now managed to resumed working on this toppic. In order 
> to exclude basic errors or errors introduced by third party RPM's I started 
> from scratch on both a i386 and a amd64 system, both SMP running kernel 
> version, and both reinstalled from scratch. Both nodes are AMD Opteron nodes. 
> The operating system is SL303, and the kernel version 2.4.21-20.ELsmp. I 
> started with the original tar balls from open IB (version openafs-1.3.80), 
> and did a basic configuration aka
> ./configure --enable-transarc-paths
> make
> make dest
> repeating the same step on both nodes. While it works on the i386 system, I  
> reproduced the reported problem on the 64bit Opteron node, that is when 
> trying to start the client without -dynroot it gets stuck (although the 
> root.afs and root.client volumes are  there) , if using -dynroot afs is 
> mounted, but if I try to access it eg. with fs, fs itself gets stuck 
> (previously I evens saw a segfault at this step but no oops in the syslog 
> which I could send you). So, I think there is definitely problem for Opterons 
> on 64bit. I can live running the box with i386 system on it, and that is most 
> probably what I'm going to do now, but if there is any more piece of 
> information that I can send to you to be able to investigate the problem 
> please let me know. It would be really nice to find a solution for this :-)
> 
> Many thank's for all the tipps that I got from you!
> Ulrich


-- 
-----------------------------------------------------------------
Hartmut Reuter                           e-mail reuter@rzg.mpg.de
					   phone +49-89-3299-1328
RZG (Rechenzentrum Garching)               fax   +49-89-3299-1301
Computing Center of the Max-Planck-Gesellschaft (MPG) and the
Institut fuer Plasmaphysik (IPP)
-----------------------------------------------------------------