ACPI Errors Causing the System to Halt During Normal Operation

[wil421]

Hi All,

Recently noticed by Mac was complaining about its backup not completing for 10 days and noticed my FreeNAS box was frozen after 119 days of uptime. No new hardware has been added and I haven't done any upgrades recently.

It look like there are some ACPI errors in the console. Image below.

ACPI Error: AE_ERROR, Thread 100002 could not acquire Mutex [ACPI_MTX_Namespace] (0x1) (20171214/utmutex-434)
ACPI Error: AE_ERROR, While evaluating Sleep State [/_s5_) (20171214/hwxface-636)
acpi0: AcpiEnterSleepStatePrep failed - AE_ERROR

 

[sretalla]

You haven't shared it (yet), but I'm going to guess that you may be using a Ryzen box.

You need to disable C states and Cool 'n' quiet in the BIOS.

 

[wil421]

No ryzen. Just an i3.

Build:
CPU: i3-8100
MB: X11SCH-LN4F
RAM: 16GB x2 of Supermicro (MTA18ADF2G72AZ-2G6E1) SDRAM ECC Unbuffered DDR4 2666
HD: 6x WD Red 4TB HDs WD40EFRX
Boot Drive: 2x 32gb Generic USB 3.0 Sticks
OS: FreeNAS-11.2-U8

Services: NFS, S.M.A.R.T., SMB, UPS
No Jails or VMs.
1 ZFS pool taking up the available space on 6 the WD Red Drives.

 

[sretalla]

No ryzen. Just an i3.

OK, so it's not that simple.

A couple of things I note from the screenshots:

Did you take 2 screenshots because this happened twice? or are you just showing information not visible in the first capture in the second one?

It seems to me that what's there is describing the end to a system shutdown (where the attempt to signal the power state to the BIOS isn't working).

What's strange about that is that you mention nothing about initiating a shutdown, so maybe we need to look into why it thinks that was a thing to do.

 

[wil421]

Yes I only took the screenshots because the NAS wasn't being responsive to the web UI and losing connection with the network drives the NAS provides. The NAS runs 24/7 unless I need to upgrade. I don't have any automation or scheduling that would cause a shutdown.

 

[sretalla]

To recap the situation...

It seems someone or something initiated a shutdown.

During the shutdown process, an attempt to tell the BIOS to go into a sleep state (or power down) failed.

The OS shutdown finished and it was telling you to press any key to reboot (or I suppose you could have hit the physical power button if you were the one initiating the shutdown).

I don't really see any information here that can point us to the "problem", which in this case is the initiating of the shutdown, not the power state failure.

 

[wil421]

Let me try to throw the circuit breaker that's connected to the UPS that powers the computer. The UPS is setup to initiate a shutdown if the battery is getting low. Its the only thing I could think of that initiates a shutdown. Only problem is I lost power last night and didn't have an issue.

 

[sretalla]

Only problem is I lost power last night and didn't have an issue.

I guess the duration of the power cut would be the important factor... longer would make it more likely to trip the threshold for shutdown.

 

[wil421]

I did some testing yesterday and it is the UPS that is trying to initiate a shutdown. My UPS uses /sbin/shutdown -p now to start the shutdown after 5 minutes, I assume this is a default script for shutdown.

When I initiate a shutdown or reset from the IPMI console the system powers off or resets. When the UPS does it the system freezes.

 

[sretalla]

I assume this is a default script for shutdown

Maybe that's right for that older version.

In newer versions it would be /usr/local/bin/midclt call system.shutdown

 

[wil421]

I've updated the script. Going to do a couple more tests when I have time today.