We had a power outage last night and when the power came back on the server's BIOS jumbled my boot order so I had to reset the USB drive as the first boot device.  It then booted up into unRAID, but it showed disk2 as missing.  I went into the devices tab and and on the dropdown list for disk2 it showed the drive that was missing so I re-assigned the drive.  Back at the main screen I get this:

 

 

Should I choose Start or Restore and what's the difference?  I would've assumed I would just choose start but because it asks if I'm sure I want to do this, it makes me a little unsure.

 

I should note that the re-assigned drive has data on it. It's not a newly added drive or anything.  The whole array has been being used and and no drives have been changed for months.

Because you re-arranged the boot order the unRAID server thinks you are upgrading disk2. 

 

If you press "Start" the array should start as expected.

 

If you press "Restore" the system.dat file that holds the superblock data for your array will be renamed to system.old.  This will force unRAID to create a new super.dat file based on the current device assignments you've made on the devices page, and then also force it to re-calculate parity on the entire array since it thinks it is starting up a brand new system.

 

The first ("Start") will just start the array back up, and since the array was not shut down cleanly, it will probably start a full parity check.

The second will immediately invalidate parity and leave your array unprotected until parity is completely re-calculated.   

In this case, both will probably have much the same result, but only because the power fail resulted in a unexpected shutdown.

 

You should invest in a UPS... it saves a lot of headaches.

 

Joe L.

Thanks.  That's what I figured.  I pretty much knew what restore would do, but the wording for Start bothered me a little and I wanted to be sure.  One question though.  Why would changing the boot order in BIOS have any effect on the drives in unraid?

 

I know I should get a UPS, but I always forget about it,  Plus I have no idea how big of one to get for two servers with 6 and 8 drives respectively and allow for expansion.  Also, isn't it currently not possible to integrate unraid with a UPS to allow for graceful shutdowns?  Our outage last night was almost two hours.  I know it's still a good idea to have one for short outages, flickers, and fluctuations.  It's on my to-do list.  n fact it just got bumped up that list. 

Well I hit Start and it's doing a Data-Rebuild!  I guess it's rebuilding disk2 from parity?  Thank goodness for the parity, but why is it acting like the data on disk2 is gone?

Well I hit Start and it's doing a Data-Rebuild!  I guess it's rebuilding disk2 from parity?  Thank goodness for the parity, but why is it acting like the data on disk2 is gone?

Apparently it thinks you replaced disk2 with a new one.

Yah, I get that.  I just don't know why.  Also, shouldn't unraid have thought my parity was "dirty" due to the improper shutdown?  Thankfully it didn't so I had parity to rebuild disk2.  This has been a bit scary because I had all my photos on that drive.  I know I should have a separate backup, but still.

Also, shouldn't unraid have thought my parity was "dirty" due to the improper shutdown?

Normally, I would think so. 

Have you installed any of the "powerdown scripts" posted here to allow a smooth powerdown?  (perhaps it caught the powerdown signal?)

Thankfully it didn't so I had parity to rebuild disk2.  This has been a bit scary because I had all my photos on that drive.  I know I should have a separate backup, but still.

Raid is not a replacement for backups.  Take 20 minutes and burn your photos to a DVD.  Keep the DVD in a different location.

 

You can interface to an APC brand UPS.  several of us have done it.  the thread describing how is here: http://lime-technology.com/forum/index.php?topic=1528.0

(you can probably interface to any UPS, but the apcupsd package for slackware makes it easy for APC brand UPS's.)

 

I'd get as big a UPS as you can manage.  I have two in my theater, one a SmartUPS 1400, the other a SmartUPS 1000.  Those cost me roughly $100 each surplus. They power my projector, receiver, and Home-Theater-PC. 

 

I have an APC  Back-UPS ES 750 on my unRAID server.  It provides about 10 minutes of run time (12 hard disks) with all disks spinning, roughly 15 minutes if they are idle.  It is connected via USB cable to the unRAID server.

 

Joe L.

Raid is not a replacement for backups.   Take 20 minutes and burn your photos to a DVD.  Keep the DVD in a different location.

 

Worth repeating: photos are items which have no natural backups, are relatively small files and DVDs store a lot of them.  Burn and store at work or at a friend's house.

 

 

Bill

Raid is not a replacement for backups.   Take 20 minutes and burn your photos to a DVD.  Keep the DVD in a different location.

 

 

I'll second that.

 

I have a 4 gig flash drive that I have all of my son's pictures on. I have that attached to a digital picture frame.

 

I also have the pictures stored on my unRAID server & on several DVD's.. I would hate to lose any of them. They cannot be replaced.

 

Phil

 

Yah, I get that.  I just don't know why.  Also, shouldn't unraid have thought my parity was "dirty" due to the improper shutdown?  Thankfully it didn't so I had parity to rebuild disk2.  This has been a bit scary because I had all my photos on that drive.  I know I should have a separate backup, but still.

 

When you press the special checkbox to start the array, unRAID gives a message on the far left under the drive table telling you something about what it is going to do to start the array.  If you had noticed it, I believe it would have said something about rebuilding the drive.  On the other hand, if it had throught that the drives were just reordered it would have said it would reorder the drives.

 

When your computer rebooted and scrambled the drive settings in BIOS, and you had to reset them, I believe it caused the BIOS to assign them in a different order.  Although unRAID is pretty good about handling this situation, it is not perfect.  Once unRAID is booted and a drive is missing, it is somewhat impossible to put the drive back in place and have unRAID "trust" it as the right drive.  It will do exactly what it is doing - and that is rebuild the drive.  If multiple drives had been unrecognized, you would not have had an ability to start the array.

 

unRAID has a somewhat nasty habbit (maybe it is Slackware's fault and not unRAID's) of NOT flushing buffers to the disk for very long periods of time (days to weeks!).  The command "sync" can be manually run after a long disk operation to force all updates to be written.  I would recommend looking at any data recently written to the array (especially the rebuild drive) and ensure it is intact and complete.

 

There is a special command that you can use in combination with the restore button to cause unRAID to "trust" your array.  It still triggers a full parity CHECK, but it trusts your data drives (which is likely what you would have preferred)  If you had noticed it was going to rebuild disk 2 and you didn't want that to happen it likely would have been suggested (or I would be suggesting it now).  You can read about it on the "Best of the Forums" (see link in my sig), "Hail to the chief" section (invalid slot nn).

 

Hope the rebuild worked okay.  I expect it did and you are fine.

unRAID has a somewhat nasty habbit (maybe it is Slackware's fault and not unRAID's) of NOT flushing buffers to the disk for very long periods of time (days to weeks!).  The command "sync" can be manually run after a long disk operation to force all updates to be written.

 

I do not believe this to be true, If it were, we would have much more corruption then we presently see.

What is left when you do a sync is updates to the superblock, regardless of that,  the data it self is scheduled and flushed at a frequent interval.

In fact the interval (according to unRAID's /etc/sysctl.conf) is more often then a regular environment.

by convention, dirty data can be in the buffer cache for 30 seconds before being written. if it's older then that it is flushed.

 

I would recommend looking at any data recently written to the array (especially the rebuild drive) and ensure it is intact and complete.

 

I do agree with this advice.

I probably didn't say this very well.  Please forgive any technical inaccuracies below, as I seem to always get myself into trouble when I try to be too technical about how Linux works. 

 

After a power failure or other dirty shutdown, it is common to have some parity check errors.  It is also common to see that a number of transactions played back.

 

See THIS thread.

 

Running "sync" forces the pending transactions to be written, and (I think) would eliminate (or at least greatly reduce) the number of parity check errors.

 

So the data IS written to the disk from the buffers, it is just not written to the eventual home - it is written to a transaction area (called the journal area, I think).  Transactions can sit in this journal area for a very long time if there is not much activity on the array (unless you run the "sync" command).  Data drives (ReiserFS) recover quite well from dirty shutdown, but somehow unRAID parity does not.

 

Bottom line, I would not want to do a drive rebuild after a dirty shutdown unless the data drive had actually failed.