Disable unclean shutdown parity check.

[Lloyd Quenby]

Hi all,

 

Is there a way to disable the parity check after an unclean shutdown?

 

The reason for this is that i have my server in my living room as i live in a small flat and due to the noise of the fans in my 5in3 hdd cases i replaced them with quieter ones. While these fan do a fine job of keeps temperatures under control in general use, during a parity check i have to put a desk fan in front of the server or the drives rapidly heat up.

 

A few days ago i came home to find out that my girl friend has turned the server on and off at the plug socket as she was having problems watching media from it, i checked the drives and they were at 45c.

 

If there is not currently a way to do this i think it would be a good idea to be able to disable this check until a more appropriate time, but to ensure it is not ignored, maybe start the server with the drives in maintenance mode so no more changes can be made to disrupt the file system.

 

Lloyd

[MyKroFt]

or my solution....

 

at my house, nobody but me touches anything on the unRAID servers - period!

 

Too much time, money and energy invested for someone to do something stupid and damage data and or hardware.....

 

I would move the server somewhere so the fan noise would not be a problem and install proper fans.  I use the fan control plugin so my fans slow down/speed up as required.

 

Myk

 

[kizer]

If I was you I'd make sure you have the Powerdown Package plugin installed. Show her how to shutdown the server using it. Would I recommend using it all the time simply because its annoying? NO, but it is/was designed to shut down the server safely. OR give her web access and show her the proper way of turning things on/off. My misses doesn't touch my server at all and just informs me when things don't seem right and I just deal with it.

 

The unclean parity check is by design. Your server doesn't know if it was writing or reading when the power was yanked so its attempting to make sure everything is in sync when it returns. You could simply stop it each time, but honestly you could be asking for trouble.

 

For the fans I currently run a 7volt mod so my fans spin a little slower in my cages, but they still seem to keep my drives cool. I used the guide below, but modified a Y-Cable so I can just plugin/remove if I need to. Honestly I've never touched it since I moded it, but its an idea just the same.

http://www.overclockers.com/forums/showthread.php/372297-Guide-to-the-7v-Fan-Mod

 

You could try something like that or install a Fan controller aka hardware knob and when you know your going to do a parity check crank up the knob so the fans can run full speed and low when you don't want to hear them, but want some air movement. Keeping the lady friend from turning the knob will be your challenge, but its an option.

 

 

[itimpi]

A possibility is to turn of the auto-start of the array.  That way you control when the array starts (and thus any potential parity check starts).

[kizer]

You could also post up a discussion in Hardware and maybe get a little advice on your Fan problem. Sounds like you might not be getting enough flow through the case.

[Kir]

...

The unclean parity check is by design. Your server doesn't know if it was writing or reading when the power was yanked

...

If all drives are spun down, it's clean. State of it could be written to flash.

Just an idea.

[JonathanM]

...

The unclean parity check is by design. Your server doesn't know if it was writing or reading when the power was yanked

...

If all drives are spun down, it's clean. State of it could be written to flash.

Just an idea.

What about kernel cached data? There is a good reason to issue a sync command before shutting down the array. I suppose you could pro-actively sync the array every time before all drives are spun down.

 

 

[trurl]

...

The unclean parity check is by design. Your server doesn't know if it was writing or reading when the power was yanked

...

If all drives are spun down, it's clean. State of it could be written to flash.

Just an idea.

What about kernel cached data? There is a good reason to issue a sync command before shutting down the array. I suppose you could pro-actively sync the array every time before all drives are spun down.

Of course none of these "improvements" would have helped with the situation in the OP anyway.

[kizer]

Nope, none.

 

The power down needs to be clean simple as that.

[Lloyd Quenby]

So in a nut shell, not possible and probably never gonna be. 

 

I have told her before not to power it off, but i think i may have to look at better cooling options as i guarantee its gonna happen again.

 

Cheers all.

[archedraft]

I have told her before not to power it off, but i think i may have to look at better cooling options as i guarantee its gonna happen again.

 

Simple solution: High voltage line to power button and power cord... it will only take one time to teach her never to touch the electronics again... 

 

[Kir]

Girlfriend upgrade time? 

[trurl]

So in a nut shell, not possible and probably never gonna be. 

 

I have told her before not to power it off, but i think i may have to look at better cooling options as i guarantee its gonna happen again.

 

Cheers all.

You should try to fix your cooling issues regardless.

[PeterB]

So in a nut shell, not possible and probably never gonna be. 

 

I have told her before not to power it off, but i think i may have to look at better cooling options as i guarantee its gonna happen again.

 

Cheers all.

You should try to fix your cooling issues regardless.

 

Indeed - I have posted on several occasions about my cooling set up - basically, remove the fans from the 3x5 cages, install multiple case fans all extracting, and block all holes in the case (other than the fan apertures) to ensure that all incoming air flows through the drive cages.  My server operates in an ambient temperature which is usually in the low 30s Celsius - drive temperatures rarely exceed 40C (during parity ops, they might reach 42-43).

 

PS. I do use the Dynamix Auto Fan Control plugin in order to regulate the fan speeds.

[SSD]

A possibility is to turn of the auto-start of the array.  That way you control when the array starts (and thus any potential parity check starts).

 

This is the answer to the question. Disabling auto array start puts you in control of when the parity check starts. It also means that media will not play until someone who knows what they are doing starts the array. So next time the gf powers down the server, she'll have to wait for you to watch anything, a definite deterrent to doing it again.

 

I have my array permanently configured this way, not because of people accidentally powering it down, but I want to know that it happened and be there when the parity check starts. Besides power loss, it is extremely extremely rare.

 

Regarded to the heat, 45C during parity check is not awful. It is sort of the teetering point between a little too warm and moderately too hot. If you typically supplement cooling during parity checks and do the tweak above, yet once in a great while the server runs at this temp for a half day, I wouldn't worry. But if temps are peaking much above that I'd look to supplement cooling.

[JonathanM]

I have my array permanently configured this way, not because of people accidentally powering it down, but I want to know that it happened and be there when the parity check starts. Besides power loss, it is extremely extremely rare.

This, exactly. I have several dockers and a couple VM's that I want to control exactly what and when things start up. I want to make sure the array is up and healthy before I start any processes that write to it, so everything is manually triggered. Also, I want things to start up in a precise order, and let things settle before moving on to the next item. If you have everything automatically firing up at once, it's chaos if something goes wrong right in the middle of the sequence.

 

Full shutdowns happen so rarely that it's no big deal to do it this way, and it makes troubleshooting and error recovery so much easier.