Nut upsuser - Jan 2014 - Shutdown when both (all) UPS on battery / low battery

On Jan 21, 2014, at 9:24 AM, Rafal Oleszek wrote:
> I tried all options:
> 1,2
> and in MINSUPPLIES: 1,2,3,4
> still didn't work.
My mistake - I gave you the wrong advice before. For MINSUPPLIES, I'm used
to either a 1:1 UPS-to-server ratio, or 2:1 where the server can run with either
power supply unpowered.

I just tried this in upsmon.conf:

MONITOR dummy1 1 monuser password slave
MONITOR dummy2 1 monuser password slave
MINSUPPLIES 1

With only one dummy-ups set to "OB LB", upsmon kept polling. When both
were set to "OB LB", it tried to shut down the system.

That was with NUT 2.7.1 - I don't remember any bugs in this area, but it
would be in the changelog.
> UPS_251 and UPS_252 feeds all environment and it's not feed directly
one
> server, but many of them.
Right, and only the master is responsible for shutting down the UPS itself. Or
do you mean that not all of the servers are connected to both UPSes?

-- 
Charles Lepple
clepple at gmail

> On Jan 21, 2014, at 9:24 AM, Rafal Oleszek wrote:
>
>> I tried all options:
>> 1,2
>> and in MINSUPPLIES: 1,2,3,4
>> still didn't work.
>
> My mistake - I gave you the wrong advice before. For MINSUPPLIES, I'm
used
> to either a 1:1 UPS-to-server ratio, or 2:1 where the server can run with
> either power supply unpowered.
>
> I just tried this in upsmon.conf:
>
> MONITOR dummy1 1 monuser password slave
> MONITOR dummy2 1 monuser password slave
> MINSUPPLIES 1
>
> With only one dummy-ups set to "OB LB", upsmon kept polling. When
both
> were set to "OB LB", it tried to shut down the system.
>
> That was with NUT 2.7.1 - I don't remember any bugs in this area, but
it
> would be in the changelog.
>
>> UPS_251 and UPS_252 feeds all environment and it's not feed
directly one
>> server, but many of them.
>
> Right, and only the master is responsible for shutting down the UPS
> itself. Or do you mean that not all of the servers are connected to both
> UPSes?
>
> --
> Charles Lepple
> clepple at gmail
>
UPS_251 and UPS_252 feeds all servers (about 40).

>On Jan 21, 2014, at 9:27 AM, Rafal Oleszek wrote:
>
>> So it's impossible to configure upssched to start timer only if all
ups
>> are in status: OB or LB ?
>
> Not impossible, just not possible inside a single configuration file.
You > can always
> put that logic into an external script called by upssched.
NUT is installed only on one server and scripts (to order shutdown
environment) are called by upssched when UPS_251 and UPS_252 change state
into OB (first script) and LB (second script). But NUT isn't waiting for
both UPS. Just start timer when only one goes on OB. Maybe it is working
only for SHUTDOWNCMD?
In manual there is a section about timers: * mean that if ANY of ups goes
OB it start proceed. Maybe that why it's not waiting for both UPS?

On Jan 23, 2014, at 2:33 AM, Rafal Oleszek wrote:
>> On Jan 21, 2014, at 9:27 AM, Rafal Oleszek wrote:
>> 
>>> So it's impossible to configure upssched to start timer only if
all ups
>>> are in status: OB or LB ?
>> 
>> Not impossible, just not possible inside a single configuration file.
>> You can always put that logic into an external script called by
upssched.
> 
> NUT is installed only on one server and scripts (to order shutdown
> environment) are called by upssched when UPS_251 and UPS_252 change state
> into OB (first script) and LB (second script). But NUT isn't waiting
for
> both UPS. Just start timer when only one goes on OB. Maybe it is working
> only for SHUTDOWNCMD?
Correct, SHUTDOWNCMD in upsmon is considering MINSUPPLIES and the powervalue
numbers. By manipulating MINSUPPLIES, you can get either a Boolean AND or OR of
all the LB states.

NOTIFYCMD (which triggers upssched) is only triggering on one UPS at a time.
> In manual there is a section about timers: * mean that if ANY of ups goes
> OB it start proceed. Maybe that why it's not waiting for both UPS?

Yes, but you can start two timers, and wait for OB status on both UPSes using
logic in your upssched CMDSCRIPT script.

I admit that in your setup, there appears to be some asymmetry between upsmon
and upssched when handling OB/LB states. However, the LB case does not have the
added complication of timers. One reason I tend to not rely on OB for anything
other than notifications is that the UPS calibration routines should track any
decrease in battery runtime when triggering an LB event. If the battery runtime
were to fall below the timer values, the system would have an uncontrolled
shutdown.

If your UPS supports adjusting the LB threshold, I would consider raising
battery.charge.low and/or battery.runtime.low, and starting the shutdown
sequence at that point. (In contrast to the battery runtime, your shutdown
procedure times will likely not change unexpectedly over the long term.)

-- 
Charles Lepple
clepple at gmail