Am 27.01.23 um 12:10 schrieb Bj?rn Baumbach:> Hi Stefan,
>
> On 1/27/23 08:57, Stefan Kania via samba wrote:
>> Am 26.01.23 um 22:57 schrieb Martin Schwenke:
>>> winbindd and smbd are only started during the CTDB startup event,
so
>>> you have to enable the event scripts before you start CTDB.
>>>
>> That's what I did and that's where the problem starts ? As soon
as I
>> enable one of the two events the cluster chrashes ?
>
> you described, that you startup the cluster and wait until the nodes are
> "OK". Then you enabled the winbind script, which detects that the
> winbind is not running, because it is running in the "monitor"
mode. The
> node goes "UNHEALTHY". This is expected.
> The monitor script does not try to start the winbind service in this
> case (monitor mode).
>
> CTDB only tries to start the enabled services when the node is in the
> "startup" run state.
>
> Best regards,
> Bj?rn
>
I just checked it in a test environment: It doesn't matter if I start
ctdb first then enable the event 49.winbind and 50.samba then restart
ctdb or if I first enable the events and then start ctdb. In both cases
ctdb will start and it's working.
The only difference between my test environment and the production
system is, that I have two server with Ubuntu and one with Debian, BUT
both with the same sernet-packages, and the same nodes- and
public_addresses-files. Still ctdb is running IP-failover works, but as
soon I enable one of the events there is no way to get ctdb restarting
again.