registry config not being updated on one of my clusters ... how do I troubleshoot this

[Richard Sharpe]

On Wed, Dec 10, 2014 at 12:22 PM, ronnie sahlberg
<ronniesahlberg at gmail.com> wrote:
> On Wed, Dec 10, 2014 at 2:58 PM, Richard Sharpe
> <realrichardsharpe at gmail.com> wrote:
>> On Wed, Dec 10, 2014 at 11:39 AM, ronnie sahlberg
>> <ronniesahlberg at gmail.com> wrote:
>>> Maybe you are running a different smbd binary on that node?
>>>
>>> Check with 'ls -l /proc/<smbd-pid>/exe' on both nodes.
>>>
>>> This should give you the exact path to the binary that is running.
>>> Verify that this is the same path on both nodes   and then check with
>>> md5sum on the two nodes and verify that they are exactly the same.
>>
>> The md5 sum is the same on both nodes, but it is also the same on the
>> two known working nodes.
>>
>> Back to single stepping the code to see what is going wrong.
>>
>> At some point it should contact ctdb I imagine when it opens the registry file.
>
> Yeah, or a little before.
>
> I think should be contacting ctdb somewhat before it opens the registry file.
> When running in clustered mode it gets the paths to use for all the
> tdb files from the ctdb daemon.
>
>
> A final quick question,    did you restart smbd after you had enabled
> clustering in the config file
> or might it be that the main smbd daemon that is running is an
> instance from before you setup ctdb ?

Hmmm, not sure. On the other cluster I did not have Samba running
before this. On this one I might have.

Anyway, changing the smb.conf to this:

[global]
clustering = yes
include = registry
config backend = registry

has made things work the way I think they should. That is, adding the
'include = registry' seems to have done it. When I was in gdb I
noticed that lp_clustering returned false all three times it was
called on one node when running net conf list, but returned true on
the last call on the first node.

Strange.

-- 
Regards,
Richard Sharpe
(何以解憂？唯有杜康。--曹操)