Odd WINS Problem

[Peter Samuelson]

[William R. Knox <wknox at mitre.org>]
> One other odd piece of behavior - when it happens, it seems to affect
> ALL of our samba servers registered with the same WINS server at once
> (or at least within a matter of hours - we know when it happens
> fairly quickly, but not instantly), while the samba servers
> registered with another WINS server are unaffected.

Hmmmm.  So we have either

a) subtle NT bug in which Samba protocol data, though valid, is not
identical to what Windows sends so it is somehow misinterpreted,

b) subtle Samba bug of not responding correctly to some quirk in the
documented protocol,

or

c) evil Microsoft plot to discredit Samba.  (Think "Oh look, Windows
won't run on DR-DOS".)

My money is on (b).  I unfortunately know basically no technical
details of NBNS protocol or SMB in general for that matter.

If possible, you might crank up the debug level for nmbd and see if you
catch any anomalies.  Start it with `-d 50' or so.  Yes, that may mean
having to rotate log.nmb more frequently than usual....

> The only solution we've come up with is to stop and start the nmbd
> process every now and again in order to force it to reregister with
> the WINS server. But that smacks of the maintenance reboots that I
> hear NT admins have to do ;-)

If you ever have to explain this one to them, ask them if their
maintenance reboots cause a downtime of longer than ~ half a second.
If so, I think you still have the edge. (:

Peter