[torqueusers] pbs_iff errno 15007 again

[Garrick Staples]

On Mon, Jan 14, 2008 at 06:22:08PM +0100, Wilhelm Eger alleged:
> Hi there,
> 
> I have a HP UX node as a torque node, which gives me the following for
> qstat/pbsnodes etc.:
> 
> pbs_iff: Invalid credential MSG=cannot authenticate user
> No Permission.
> qstat: cannot connect to server xxxx (errno=15007)
> 
> This appears periodically. Sometimes all seems to be right (e. g. right
> output from qstat). The periods are about 2 to 20 seconds.
> 
> I can't set up a job an this node either. pbs_iff is in the path and is
> setuid.
> 
> pbs_iff -t xxx 15001 gives no error. I guess it depends on the  NIS system.
> The node ran before i rebooted it this day. I have two nodes, which are HP
> maschines,too, and they're running smoothly. I know, there is something with
> NIS and pbs_mom.
> 
> 
> Could anyone give me an hint?

In src/lib/Libifl/pbsD_connect.c, you'll find the following:
/* FIXME: is this necessary?  Contributed by one user that fixes a problem, 
   but doesn't fix the same problem for another user! */
#if 0
#if defined(__hpux)
   /*HP-UX : avoiding socket caching */
   send(connection[out].ch_socket, '?', 1, MSG_OOB);
#endif
#endif


Can you change that 0 to a 1 and try it?

-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
Url : http://www.supercluster.org/pipermail/torqueusers/attachments/20080114/7b93da2a/attachment.bin