[Samba] CTDB and locking issues in 4.4.6 (Classic domain)

Volker Lendecke vl at samba.org
Thu Nov 3 19:37:05 UTC 2016

On Thu, Nov 03, 2016 at 07:11:12PM +0000, Alex Crow via samba wrote:
> On 03/11/16 14:28, Ralph Böhme wrote:
> >
> >> Ah, my smb.conf is missing the "ctdb socket" parameter. According to "man
> >> smb.conf" this should be set, but I don't remember seeing it on the wiki
> >> pages.
> > if it's not set, the default is used which works in many cases, but
> > not in mine. If the default wasn't working in your setup, you'd
> > notice, as clients wouldn't be able to talk to ctdb at all.
> >
> > -slow
> I can confirm this made no difference. I see the locking working from
> windows between the first two servers but *not* the third. This seems to
> match what I had with smbclient. Again, could it be that an FS that
> passes ping_pong not be able to work in a CTDB cluster?
> The developers of my fs (MooseFS) have said that while file flock() is
> working perfectly there is still no directory flock. Could that be a
> pointer?

Does your system have consistent inodes and device ids? Does

stat /some/shared/file

give exactly the same information on all nodes, in particular the
Device/Inode fields?


More information about the samba mailing list