nut problem

[maxbet]

Friday evening we had a blackout and I found all equipment shut down and the UPS with battery low (as expected). Once I powered up the system I got an email stating that the FreeNAS system has not been shut down properly.
My FreeNAS server is directly connected to the UPS (see sig) via USB and is acting as master system, while a pfSense machine and an Ubuntu server are both configured as slave. This setup has been working flawlessly for a few years now, and in previous instances all machines shut down properly and there was no warning after the successive power up.

I decided to investigate some further.

The nut log for the blackout event shows the following :

I simulated again a blackout by pulling the plug from the UPS and waited for the UPS to perform the shutdown. Strange enough, after the programmed time, the slave systems did shutdown properly but the FreeNAS server just sat idly without any message whatsoever (I was logged in with a SSH shell to monitor the shutdown message); the machine was not responsive and I had to shut it down by pressing the power button.

The log file for such an event was :

Thinking of a problem in the shutdown command, I restarted the system and tried to type in the CLI the same command programmed for nut and the machine shut down correctly; this should rule out HW support issues but left me with the doubt of user right issues. I checked with the gui and all nut process are running under the uucp user :

I checked the uucp user and it is part of operators group, as indicated in FreeBSD guidelines.

My nut setup is the following :

In my Ubuntu server, the nut processes do run under root operator.

As said above, this setup has worked correctly for a few years now, what can I check in order to debug this problem? Any ideas?

Max

 

[Elliot Dierksen]

As said above, this setup has worked correctly for a few years now, what can I check in order to debug this problem? Any ideas?

Is it possible that your battery discharges prior to the 300 second delay time expiring? If the slave servers shut down immediately on a power fail, perhaps you could decrease the delay time. I have an APC power chute appliance shut down the VM hosts. FreeNAS is configured to shutdown immediately on a low battery condition.

 

[maxbet]

I would rule out hardware related problems.
The test running the command from the shell and getting a proper shutdown, proved that the hardware supports correctly the shutdown.
The simulation done has shown that the UPS broadcasted the shutdown message when it was at 75% of battery level, therefore there was more than enough energy to complete even a 5 mins long shutdown.
The issue is/was that after broadcasting to the slaves, FreeNAS stood idle and did not power down. This, in my opinion, is symptom of a SW issue.