samba_spnupdate invoked oom-killer ( samba BUG )

[Andrew Bartlett]

On Sat, 2014-10-04 at 11:50 +0200, Volker Lendecke wrote:
> On Thu, Oct 02, 2014 at 06:56:04AM +1300, Andrew Bartlett wrote:
> > On Wed, 2014-10-01 at 17:35 +0200, Volker Lendecke wrote:
> > > On Wed, Oct 01, 2014 at 08:33:13AM -0700, Jeremy Allison wrote:
> > > > On Wed, Oct 01, 2014 at 06:44:51AM +0200, Volker Lendecke wrote:
> > > > > On Tue, Sep 30, 2014 at 11:46:24AM -0700, Richard Sharpe wrote:
> > > > > > On Tue, Sep 30, 2014 at 10:24 AM, Chan Min Wai <dcmwai at gmail.com> wrote:
> > > > > > > Hi Richard,
> > > > > > >
> > > > > > > I think smbcontrol didn't work on samba daemon (the AD DC daemon)
> > > > > > 
> > > > > > Ahhh, this is an AD DC problem. My experience is only with the file
> > > > > > server side of the code. I think you need other people to check in
> > > > > > here to figure this out.
> > > > > > 
> > > > > > I seems like there is a leak. I would increase the log level to see if
> > > > > > I can see anything that is happening over and over again ...
> > > > > 
> > > > > FYI: In the last weeks I've got smbcontrol working against the DC,
> > > > > so in the near future we'll have the infrastructure to do a debug or
> > > > > pool-usage message against the DC.
> > > > 
> > > > W00t! Great work Volker, thanks. This is something that
> > > > has been sorely needed and will greatly increse reliability.
> > > 
> > > Well, if I only had time now to get this upstream :-(
> > 
> > Is this something I can help with?
> 
> You might want to take a look at the attached patchset. This
> is what I've dug out of my huge messaging mess. It compiles,
> but right now I don't have the time to even do some simple
> tests. At least that's the direction where I want to go
> eventually.
> 
> Maybe you already have some comments.

Looking over them quickly, I did want to re-assure you that putting the
messaging sockets in the lock directory is quite reasonable, as I see
it, because like other lock files, they may/should be removed by the OS
on reboot.  Perhaps put both the privileged and public sockets under the
lock dir?

(That said, there isn't any practical or philosophical reason the cache
dir it isn't used, either just accidents of history or quite possibly
there hasn't been any caches in that code yet!)


> Thanks,
> 
> Volker
> 
> 
> Maybe you already have some comments.
> 
> Thanks,
> 
> Volker
> 

-- 
Andrew Bartlett                       http://samba.org/~abartlet/
Authentication Developer, Samba Team  http://samba.org
Samba Developer, Catalyst IT          http://catalyst.net.nz/services/samba