C2SEA won't ACPI power down

[JonathanM]

I've been successfully running unraid for a couple years now on a Supermicro C2SEA, and everything except ACPI shutdown works fine. I've used 4.4 and 4.7, and they both act the same, the screen shows the last thing as a shutdown being requested, but it doesn't ever power down.

 

Anyone else run into something similar?

 

Jonathan

[dabl]

Yes!  And I've been meaning to post about it.

 

Super annoying and it's hard to imagine it's supposed to be this way and/or somebody else hasn't run into it.

 

I found that on my board it would only work if I turned off the setting for Restore on AC Power Loss to Last State (and maybe Power-Off) but NOT the one I obviously would want (Power-On)

 

Would be interested to see if you see the same behavior.

 

[JonathanM]

Could you please check your syslog for a line like this? I think it may be related.

 

ACPI Warning: Incorrect checksum in table [OEMB] - 7D, should be 7C (20090903/tbutils-314)

 

It's early in the boot process, during the hardware discovery phase.

[dabl]

Could you please check your syslog for a line like this? I think it may be related.

 

ACPI Warning: Incorrect checksum in table [OEMB] - 7D, should be 7C (20090903/tbutils-314)

 

It's early in the boot process, during the hardware discovery phase.

 

I haven't rebooted in quite awhile so I don't have boot related entries in my current log.

 

What were your results from the settings I suggested?

[JonathanM]

 

I haven't rebooted in quite awhile so I don't have boot related entries in my current log.

 

What were your results from the settings I suggested?

Like you said, I too have not had a chance to reboot. This is a production machine, it stays up for months unless there is a power failure or other event. Do you not have any saved logs?

[dabl]

Like you said, I too have not had a chance to reboot. This is a production machine, it stays up for months unless there is a power failure or other event. Do you not have any saved logs?

 

I checked the previous logs and I do have this entry.  I'm still running v4.6 need to update to v4.7 and will check for that.  Also my model is a C2SEE.

 

The real question is what (if anything) it means and/or whether the many other users with this motherboard see this behavior etc.

[JonathanM]

Setting the ACPI to "Last State" allowed it to power down correctly. So our systems do indeed act exactly the same way. What BIOS revision do you have in yours?

[dabl]

Setting the ACPI to "Last State" allowed it to power down correctly. So our systems do indeed act exactly the same way. What BIOS revision do you have in yours?

 

Can't look at the moment but I'm sure it's 1.0c based on what's at the website.

 

I checked to make sure mine was up to date when I bought it and it looks like nothing has changed since I last checked. 

 

It does seem odd nobody else seems to have complained about this behavior since this is the default unRAID motherboard......

[JonathanM]

I checked to make sure mine was up to date when I bought it and it looks like nothing has changed since I last checked. 

 

It does seem odd nobody else seems to have complained about this behavior since this is the default unRAID motherboard......

 

Darn. I guess updating the BIOS won't help then.

 

Perhaps nobody else changed the power on state from default?

[dabl]

Now that you mention it I guess I haven't tested this thoroughly.

 

I was just glad to get the freaking thing to power down PERIOD.

 

I don't know or care what the 'default' is, when power fails I want the server to power back on when it returns.

 

Always.  

 

To me that does NOT mean 'last state'.

 

It's incomprehensible to me that all the geeks on this site running ups backed systems with the default motherboard haven't tested or run into this issue but since we're apparently two guys whistling in the dark here.....

 

Hello?

 

Is there anybody out there?

 

Tom?

 

 

[Rajahal]

Sorry I'm late to this thread.  Let me make sure I understand the issue.  By ACPI power down, do you mean powering down from the button on the unRAID web interface, by the power button on the server, or by some other method?  Are you using a UPS control the server's power states?

 

I've used the C2SEE motherboard extensively and I've never had any issues powering it down from the unRAID web interface or the power button or the powerdown script.  I have not tested the board with a UPS, so I can't comment there.  However, I believe that the APCUPSD script calls up the powerdown script, so I don't see any reason why it wouldn't work.

 

I am of the opinion that the power on state in BIOS should never be set to 'power on' or 'last state' for a server, but always 'power off'.  I make this change in the BIOS of every server I build.  My reasoning is that power failures tend to come in waves - you lose power for a minute, then it comes back on for a few minutes, then you lose it again, etc.  Every time the server loses power, it has to run a parity check the next time it is turned on.  So by using 'power on' or 'last state' you are allowing your server to run a parity check for a few minutes, then lose power again, then restart the parity check for a few minutes, then lose power again, etc.  It just doesn't seem like a good idea to me.  If the server is going to go down hard, I would rather it go down once and then give me a chance to wait out the power outage before manually turning it back on again.

 

Of course the best solution is to use a UPS and run APCUPSD on the server so that it can shut down gracefully in the event of a power outage.

[JonathanM]

All my machines are on online (not switching) battery backups with 45+ minutes of backup power. When the power fails, I want the server to ride out short outages (less than 10 minutes) and then commit to a full powerdown. When the power is restored, I want the server to come back up automatically.

 

And yes, by ACPI powerdown, we mean ANY software powerdown event fails. The server just sits there and looks at you, with a powerdown event requested as the last line on the terminal screen. You have to pull the plug or hold down the power button to shut the machine off. It works ok like that, it just drains the battery faster. It would be nice to be able to command a full shutdown from the command line or a script.

[Rajahal]

I see.  So you are saying the community powerdown script fails as well, even when invoked manually (i.e. typing 'powerdown' at the system console or via telnet)?  That is odd, and not something I've seen before.

 

Specifically what make and model of UPS do you have?  Does it connect to the server via USB or serial?  Can you post your settings for the APCUPSD script?  Maybe something in there is configured incorrectly.  Post a syslog also - it will contain a message if the APCUPSD script is failing in any way.  For example:

 

May 27 15:20:55 ServerPop apcupsd[2018]: apcupsd FATAL ERROR in linux-usb.c at line 609 Cannot find UPS device -- For a link to detailed USB trouble shooting information, please see <http://www.apcupsd.com/support.html>. (Errors)
May 27 15:20:55 ServerPop apcupsd[2018]: apcupsd error shutdown completed (Errors)

 

Are you running any other add ons that might be preventing the server from powering down?

 

Perhaps you can post all of your BIOS settings?  Or if that is too much work, perhaps you could restore your BIOS settings to default and start over.  Set your drives to AHCI, choose the USB flash drive as your boot drive, and change nothing else.  See what happens....

 

I'm still against setting the BIOS to power your server on automatically, I believe that is a bad practice, but it is your decision.  As far as troubleshooting this problem, I would suggest leaving it at the default setting.

[JonathanM]

I see.  So you are saying the community powerdown script fails as well, even when invoked manually (i.e. typing 'powerdown' at the system console or via telnet)?  That is odd, and not something I've seen before.

 

Specifically what make and model of UPS do you have?  Does it connect to the server via USB or serial?  Can you post your settings for the APCUPSD script?  Maybe something in there is configured incorrectly.  Post a syslog also - it will contain a message if the APCUPSD script is failing in any way.  For example:

 

May 27 15:20:55 ServerPop apcupsd[2018]: apcupsd FATAL ERROR in linux-usb.c at line 609 Cannot find UPS device -- For a link to detailed USB trouble shooting information, please see <http://www.apcupsd.com/support.html>. (Errors)
May 27 15:20:55 ServerPop apcupsd[2018]: apcupsd error shutdown completed (Errors)

 

Are you running any other add ons that might be preventing the server from powering down?

 

Perhaps you can post all of your BIOS settings?  Or if that is too much work, perhaps you could restore your BIOS settings to default and start over.  Set your drives to AHCI, choose the USB flash drive as your boot drive, and change nothing else.  See what happens....

 

I'm still against setting the BIOS to power your server on automatically, I believe that is a bad practice, but it is your decision.  As far as troubleshooting this problem, I would suggest leaving it at the default setting.

I just recently installed the apcupsd power down control, up until a few months ago I had a box stock install, and it wouldn't power down from the stock web interface either. I recently added the power button and other scripts, none of those power down either. The only way to get a software power down is to set the BIOS to last state. It refuses to kill the power if it's set to power on.

 

I'm intrigued by your statement that it's bad practice to allow a server to power on automatically when power is applied, could you give me a scenario using my current setup that would result in a bad situation? I can't think of one, unless you assume my battery backup fails, in which case I'm hosed anyway. A short (less than 10 minute) power failure results in no shutdown, and 30+ minutes reserve battery left. A longer power outage results in a controlled shutdown. If power resumes before the battery backup runs out of power, the server stays down, as it never lost power. If the power stays out long enough to drain the main UPS's batteries, it will restore power only after there is stable enough power to recharge the batteries, and the server will come back up automatically. Worst case scenario, I have to wake it up with a magic packet. We seldom (as in I can't ever remember one) have outages longer than 10 minutes and shorter than 1 hour. The UPS in charge of telling the machine when there is a power outage is a remote network connected unit not even powering any equipment. My high capacity online UPS isn't communicating with anything, it just sits there and provides power to the server until the batteries die.

 

I'm perfectly fine leaving it as is and working around the issue, it's just strange that nobody else seemed to have the issue except dabl and I. dabl found a solution that kinda works with "last state". Telling me to not use "Power on" as a solution to the issue isn't an answer, as I know how to work around the problem, I just wondered if there was something I could do to make the C2SEA system work as my other system with a Gigabyte motherboard does flawlessly.

 

As far as some setting or script not allowing power down, that seems to be a red herring, as the power down works correctly if that one BIOS setting is changed. No other changes, just changing from power on to last state makes everything work. My assumption at this point is that the ACPI handling in this linux build is not handling this single case correctly. If it wasn't a production server, I would try version 5 just for grins.

 

Do you have this line "ACPI Warning: Incorrect checksum in table [OEMB] - 7D, should be 7C (20090903/tbutils-314)" in your syslog?

[dabl]

Raj,

 

thanks for your input.

 

Jon and I have the same goal and setup which is totally different from yours.

 

Since we use a UPS and you don't, we wouldn't have the scenario you describe during very frequent power failures and restores.

 

We need to hear from any of the many other users with this motherboard and a ups.  Again it's hard to believe this topic hasn't been explored before.....

 

Jon, have you actually tested a power failure scenario and seen that the system doesn't power back up after a power failure with the bios set to 'last state'?  I have not (yet).  Logically I expect it will not but we should prove that.

[Rajahal]

I do use a UPS, though currently my server just runs on a 'dumb' UPS that will protect it against power blips but not extended power outages.  I plan on rectifying this situation soon.

 

This is the bit that makes me uncomfortable:

 

If the power stays out long enough to drain the main UPS's batteries, it will restore power only after there is stable enough power to recharge the batteries, and the server will come back up automatically.

 

Perhaps your UPS works differently from mine, but that is definitely not how mine works.  As soon as the mains come back on, the UPS passes power through to all devices while simultaneously charging its battery.  The more power the devices are using, the slower the battery will charge.  In this scenario, if there were another power failure soon after, the UPS may not have charged enough to get the server through a clean shut down and the server may go down hard.  I would rather manually wait for the UPS to charge for 15 minutes or so before turning on the server myself.  Perhaps this is a bit of a far-fetched scenario, I don't know.  I don't see the appeal of having the server turn itself back on since it isn't much trouble to walk over to it and push the power button.  Maybe you have your server tucked away in an attic or something...

 

If you could program the UPS to only turn the server back on when its battery is over 50% full, then that would be great.  I don't know how to do that.

[cyrnel]

This might be reaching but I don't see it mentioned. What are you guys using for kernel boot parameters? (/boot/syslinux.cfg)

[JonathanM]

I do use a UPS, though currently my server just runs on a 'dumb' UPS that will protect it against power blips but not extended power outages.  I plan on rectifying this situation soon.

 

This is the bit that makes me uncomfortable:

 

If the power stays out long enough to drain the main UPS's batteries, it will restore power only after there is stable enough power to recharge the batteries, and the server will come back up automatically.

 

Perhaps your UPS works differently from mine, but that is definitely not how mine works.  As soon as the mains come back on, the UPS passes power through to all devices while simultaneously charging its battery.  The more power the devices are using, the slower the battery will charge.  In this scenario, if there were another power failure soon after, the UPS may not have charged enough to get the server through a clean shut down and the server may go down hard.  I would rather manually wait for the UPS to charge for 15 minutes or so before turning on the server myself.  Perhaps this is a bit of a far-fetched scenario, I don't know.  I don't see the appeal of having the server turn itself back on since it isn't much trouble to walk over to it and push the power button.  Maybe you have your server tucked away in an attic or something...

 

If you could program the UPS to only turn the server back on when its battery is over 50% full, then that would be great.  I don't know how to do that.

The UPS I actually power the server with is an online model, not switching. It runs the inverter continuously, and maintains the batteries whenever there is power. That way, there is no switching transient at all, the output waveform is continuous. When the batteries are drained it won't turn on the output until the batteries recover somewhat.

 

Manually turning on the server requires getting in the car and driving to another location, thus my desire to not need a button push.

 

Jonathan

[JonathanM]

This might be reaching but I don't see it mentioned. What are you guys using for kernel boot parameters? (/boot/syslinux.cfg)

Box stock as far as I know.

 

default menu.c32

menu title Lime Technology LLC

prompt 0

timeout 50

label unRAID OS

  menu default

  kernel bzimage

  append initrd=bzroot rootdelay=10

label Memtest86+

  kernel memtest

 

[cyrnel]

Try adding reboot=force.

 

as in modify this line:

 

  append initrd=bzroot rootdelay=10

 

to be:

 

  append initrd=bzroot rootdelay=10 reboot=force

 

For testing purposes it might be best to try it from the loader screen. Hit tab while it's counting down. Changes you make there are used but not saved.