Nut upsuser - Jul 2017 - Trying to understand sdorder

My understanding is that NUT provides a mechanism to shut down multiple 
systems protected by multiple UPS's in a given order.  For example to 
shutdown the users before their NFS server.

1. man ups.conf says

   sdorder Optional. When you have multiple UPSes on your system, you 
usually need to turn them off in a certain order. upsdrvctl shuts down all 
the 0s, then the 1s, 2s, and so on. To exclude a UPS from the shutdown 
sequence, set this to -1.

2. config-notes.txt gives an example

To set the order in which your UPSes receive the shutdown commands, define 
the 'sdorder' value in your ups.conf.

         [bigone]
                 driver = usbhid-ups
                 port = auto
                 sdorder = 2

         [littleguy]
                 driver = mge-shut
                 port = /dev/ttyS0
                 sdorder = 1

         [misc]
                 driver = blazer_ser
                 port = /dev/ttyS1
                 sdorder = 0

The order runs from 0 to the highest number available.  So, for this 
configuration, the order of shutdowns would be 'misc',
'littleguy', and
then 'bigone'.

3. But in upsdrvctl.c I see

/* walk UPS table and send command to all UPSes according to sdorder */
static void send_all_drivers(void (*command)(const ups_t *))
{
 	ups_t	*ups;
 	int	i;

...
 	for (i = 0; i <= maxsdorder; i++) {
 		ups = upstable;
 		while (ups) {
 			if (ups->sdorder == i)
 				command(ups);
 			ups = ups->next;
 		}
 	}
}

These nested loops will execute in a few milleseconds, effectively 
shutting down all the UPS units at the same time rather than in a paced 
sequence.

I see nothing that ensures that all the "0"s are effectively shut down
before starting the shutdown of the "1"s.

Is there something in the sdorder logic that I am missing?

Roger

On Jul 8, 2017, at 4:20 AM, Roger Price wrote:
> 
> My understanding is that NUT provides a mechanism to shut down multiple
systems protected by multiple UPS's in a given order.  For example to
shutdown the users before their NFS server.
There are two factors at work here. 

With multiple systems, the order is enforced by waiting (up to a certain time,
anyway) for the slave systems to shut down first.

With multiple UPSes (and without the nowait flag), the shutdown commands are
executed sequentially - forkexec() blocks waiting for the driver. However, the
amount of time the driver spends sending the shutdown command (and possibly
waiting for an acknowledgement) is driver-dependent.