Nut upsuser - Jun 2016 - UPS is not turned off in slave-only configuration?

I have two hosts.

One is the ups-server. It's generator powered and can be assumed to never go
down. This is the one that all the UPSs are connected to. It has:
- the UPSs configured in upsd.conf
- a "upsmon master" user in upsd.users
- an empty upsmon.conf

The other one is server-1. Its upsmon.conf contains:
MONITOR test-apc at ups-server 1 monuser blah slave

Now, what I see is that when the UPS feeding server-1 goes LB, server-1 is
indeed shut down, but it seems that the UPS is never switched off.

This is a problem because when power is restored after shutdown, but before
the battery runs out (with no load, it continues for quite a while),
server-1 is not rebooted.

Is my configuration wrong? How can I help diagnosing this problem?

On Wed, 22 Jun 2016, Andr? H?nsel wrote:
> Now, what I see is that when the UPS feeding server-1 goes LB, server-1 is
> indeed shut down, but it seems that the UPS is never switched off.
>
> This is a problem because when power is restored after shutdown, but before
> the battery runs out (with no load, it continues for quite a while),
> server-1 is not rebooted.
I'm assuming that you have one UPS per server, and that server-1 is fed 
by UPS-1.

When server-1 shuts down, whichever server is in charge of UPS-1 should 
execute command

  upsdrvctl shutdown UPS-1

Does this happen?  Is this command called by a script or a systemd service 
unit?  Is there any trace of the action of that script/service unit in 
/var/log/messages or the systemd journal?

Roger

On Mittwoch, 22. Juni 2016 17:50, Roger Price wrote:

On Wed, 22 Jun 2016, Andr? H?nsel wrote:
> > Now, what I see is that when the UPS feeding server-1 goes LB, 
> > server-1 is indeed shut down, but it seems that the UPS is never
switched off.
> >
> > This is a problem because when power is restored after shutdown, but 
> > before the battery runs out (with no load, it continues for quite a 
> > while),
> > server-1 is not rebooted.
> 
> I'm assuming that you have one UPS per server, and that server-1 is fed
by
UPS-1.

Yes, but they are connected to ups-server, not to server-1. (Because
server-1 has no USB port.)
> When server-1 shuts down, whichever server is in charge of UPS-1 should
execute command
> 
> upsdrvctl shutdown UPS-1
> 
> Does this happen?  Is this command called by a script or a systemd service
unit?

I don't know. upsd has to do that when the slave has shut down, or not?
NUT is installed from the Debian (Raspbian actually) jessie package.
> Is there any trace of the action of that script/service unit in
/var/log/messages or the systemd journal?

I have only two UPS-related events in the syslog of ups-server:

(I started server-1)
Jun 24 01:32:58 ups-server upsd[561]: User monuser at 10.10.5.63 logged into
UPS [test-apc]
(Cut the power to UPS-1 and waited for the battery to become almost empty)
Jun 24 01:38:03 ups-server upsd[561]: Client monuser at 10.10.5.63 set FSD on
UPS [test-apc]

I was just about to get the syslog from server-1 when I noticed another
problem. server-1 now shuts down immediately after it boots, probably
because FSD is still set. (I read about it being a latch.)

How is this supposed to work? Is NUT not suitable for setups with
multiple-UPSs where they are connected to one server? Or did I configure it
wrong?