Best Way to Proceed...

[Griff1324]

Here is my situation.  The SMART status on my unraid dashboard recently showed one of my drives with an orange exclamation point.  Upon clicking on the disk attributes for that particular hard drive, it states that the Last SMART test was completed without error.  However, the attributes show a Relocated Sector count of 8280 and Reported Uncorrect of 22.  This is a Seagate 3TB drive, which are notorious for failing, that has been in service for over 4 years. 

 

I recently purchased an 8TB WD Red drive that is just about to complete one preclear pass.  This will be the largest drive in my server so it will need to replace my 5TB Toshiba parity drive.  I also have a new 3TB WD Red drive sitting on the shelf that will be used in my DVR.  I was thinking of doing the following:

 

1. Put the 3TB WD Red drive into my unraid server and simply copy the data off of the Seagate 3TB drive to the new WD Red.  I understand that this data will be at risk because it only resides on this one drive which is not part of the array.

 

2. Add the 8TB WD drive to the array and make it the parity drive; thus decommissioning the Toshiba 5TB drive as the parity drive.  I will have to read how to do this as it has been along time since I have had to replace / swap drives.

 

3. Add the 5TB Toshiba drive as a data disk to the array.

 

4. Copy the data off of the 3TB WD Red drive over to the 5TB Toshiba drive.

 

5. Remove the 3TB WD Red from the unraid server to be used in my DVR.

 

Do these steps sound like the best approach without putting my array at risk?

 

Chris

 

[phbigred]

Yup you can do as listed, or... If you think you can wait it out a day swap your 5TB for the 8 and rebuild parity. That way (assuming you aren't adding and removing a lot of data) you have a backup copy of your parity info should the drive die. One other option would be to pull your drive and let it emulate contents and copy the virtual data to an off array drive. This is a parity risk as is my first option, however the first one is a safer bet should your array croak. Any way you do this get that data off now before it dies. 

[SnickySnacks]

Save a couple steps and just do a parity swap?

https://lime-technology.com/wiki/index.php/The_parity_swap_procedure

I've never tried it before, but it's basically what you're trying to do.

Edited April 20, 2017 by SnickySnacks

[Griff1324]

1 hour ago, phbigred said:

Yup you can do as listed, or... If you think you can wait it out a day swap your 5TB for the 8 and rebuild parity. That way (assuming you aren't adding and removing a lot of data) you have a backup copy of your parity info should the drive die. One other option would be to pull your drive and let it emulate contents and copy the virtual data to an off array drive. This is a parity risk as is my first option, however the first one is a safer bet should your array croak. Any way you do this get that data off now before it dies. 

 

I can stop adding / removing data until this is resolved.  Just for clarification, using your first option, you are stating to leave the failing drive as part of the array and simply swap the parity drive for the 8TB drive.  Then rebuild the parity, with the 3TB failing hard drive still part of the array.  If the 3TB drive fails during the parity rebuild then I can stop the parity rebuild and put the original 5TB parity drive back in.  Is this correct?  If this is correct, then should I copy the data from the failed drive to another location first or is it not necessary?

Edited April 20, 2017 by Griff1324

[trurl]

14 hours ago, Griff1324 said:

Here is my situation.  The SMART status on my unraid dashboard recently showed one of my drives with an orange exclamation point.  Upon clicking on the disk attributes for that particular hard drive, it states that the Last SMART test was completed without error.  However, the attributes show a Relocated Sector count of 8280 and Reported Uncorrect of 22.  This is a Seagate 3TB drive, which are notorious for failing, that has been in service for over 4 years. 

 

I recently purchased an 8TB WD Red drive that is just about to complete one preclear pass.  This will be the largest drive in my server so it will need to replace my 5TB Toshiba parity drive.  I also have a new 3TB WD Red drive sitting on the shelf that will be used in my DVR.  I was thinking of doing the following:

 

1. Put the 3TB WD Red drive into my unraid server and simply copy the data off of the Seagate 3TB drive to the new WD Red.  I understand that this data will be at risk because it only resides on this one drive which is not part of the array.

 

2. Add the 8TB WD drive to the array and make it the parity drive; thus decommissioning the Toshiba 5TB drive as the parity drive.  I will have to read how to do this as it has been along time since I have had to replace / swap drives.

 

3. Add the 5TB Toshiba drive as a data disk to the array.

 

4. Copy the data off of the 3TB WD Red drive over to the 5TB Toshiba drive.

 

5. Remove the 3TB WD Red from the unraid server to be used in my DVR.

 

Do these steps sound like the best approach without putting my array at risk?

 

Chris

 

Unnecessarily complicated, doesn't really use unRAID the way it should be used, and at the end you will have to do a new config and rebuild parity since you removed a drive. Just do this:

1. Replace the failing 3TB Seagate data with the new 3TB WD Red and let it rebuild.
2. Replace 5TB Toshiba parity with the new 8TB WD Red and let it rebuild.
3. Replace the new 3TB WD Red data with the old 5TB Toshiba parity and let it rebuild.

At each step, be sure to double-check connections on all drives.

 

 

[trurl]

14 hours ago, SnickySnacks said:

Save a couple steps and just do a parity swap?

https://lime-technology.com/wiki/index.php/The_parity_swap_procedure

I've never tried it before, but it's basically what you're trying to do.

Parity Swap is also good. That would allow you to leave the new 3TB WD Red out of it completely.

 

There is no good reason to copy the data, except for the purposes of making a backup. You do have good backups of any irreplaceable or important data on another system, don't you?

[Griff1324]

52 minutes ago, trurl said:

Parity Swap is also good. That would allow you to leave the new 3TB WD Red out of it completely.

 

There is no good reason to copy the data, except for the purposes of making a backup. You do have good backups of any irreplaceable or important data on another system, don't you?

 

Turl,

 

If I opt to do the parity swap and the Seagate 3TB drive dies while the parity is rebuilding then will I lose any data?  The error count on this drive has increased over my last two parity checks that I do once per month.  

 

I do not have a backup of the data on my unraid server as I am currently around 30 TB.  Most of the data consists of movies, music, television shows and podcasts.  I could get the information back but it would be a huge pain in the butt to do so; therefore, I want to minimize any potential loss of data as much as possible.  All of my documents, pictures and other data that I can not afford to lose is on another server running two disks which are mirrored.  That information is also backed up online to Crashplan.  

[JorgeB]

4 minutes ago, Griff1324 said:

If I opt to do the parity swap and the Seagate 3TB drive dies while the parity is rebuilding then will I lose any data?  The error count on this drive has increased over my last two parity checks that I do once per month.  

 

The 3TB Seagate is not used during the swap, in fact you'll need to disable it for the swap to work.

[Griff1324]

6 minutes ago, johnnie.black said:

 

The 3TB Seagate is not used during the swap, in fact you'll need to disable it for the swap to work.

 

Sorry for all of the questions...I am just trying to wrap my head around this.  You are stating that I will need to disable the 3TB Seagate data drive before swapping out my current 5TB Toshiba parity drive with my new 8TB WD Red parity drive....is that correct?  To perform the parity sway, I pull the existing parity drive from the array, put it on the shelf, and replace it with a new empty drive.  If that is the case, then don't all of the data disks need to be present in the array in order to rebuild the parity on the new drive?  

[JorgeB]

Just follow the wiki instructions:

 

https://lime-technology.com/wiki/index.php/The_parity_swap_procedure

 

[JonathanM]

20 minutes ago, Griff1324 said:

I pull the existing parity drive from the array, put it on the shelf, and replace it with a new empty drive.

Nope. If you will read the parity swap procedure, it clearly states the original parity drive stays in the machine.

[Griff1324]

Thanks for all of the help guys.  I did read those instructions earlier but it was late last night and I did not realize that the steps were describing exactly what I was looking to do.  I will tackle this when I get home from work this evening.  Thanks again!

[itimpi]

Under the covers what the parity swap process does is first copy the parity from the old (5gb) parity drive to the new (8gb) parity drive and then uses all the other data drives plus the new parity drive to rebuild data onto the 5gb drive.  

 

During this process keep the problem 3Gb drive safe just in case anything goes wrong with the parity swap process (so that is available for emergency data recovery procedures).

[Griff1324]

I have attached my disk diagnostics here in case anyone wants to look at them and see if there are any other concerns before doing the parity swap.

sim-unraid-01-diagnostics-20170420-1512.zip

[SSD]

I would suggest adding the DVR drive as an unassigned device. Copy the data from the failing 3T drive for safe keeping.

 

Then do a new config, with the 8T drive as parity, the 5T drive parity as a data drive, and leaving out the 3T drive that is failing.

 

Once complete, format the 5T drive in the array, and then copy the data from the DVR drive, leaving the DVR drive to be the DVR drive. The failing 3T drive acts as a backup if the DVR dies during the copy. You can keep the failing drive as a backup if you don't already have the data backed up. It is of no other use.

[Griff1324]

Just now, bjp999 said:

I would suggest adding the DVR drive as an unassigned device. Copy the data from the failing 3T drive for safe keeping.

 

Then do a new config, with the 8T drive as parity, the 5T drive parity as a data drive, and leaving out the 3T drive that is failing.

 

Once complete, format the 5T drive in the array, and then copy the data from the DVR drive, leaving the DVR drive to be the DVR drive. The failing 3T drive acts as a backup if the DVR dies during the copy. You can keep the failing drive as a backup if you don't already have the data backed up. It is of no other use.

 

Correct me if I am wrong.  If I do as you stated above and do a new config....then if a drive fails when parity is being built on the 8TB drive then I lost all of that data on the failed drive since I do not have a parity disk that is complete, correct?  

 

Wouldn't I be better off copying the data from the failing 3TB drive to my DVR drive that is unassigned and then proceeding with the parity swap instructions?  That way if I suffered a failed hard drive during the parity swap, I would have an opportunity to put the failing 3TB disk back in and attempt to rebuild the failed disk using the original parity drive.  Is my logic right?  

[SSD]

Excellent question! And you are right, there might be an argument for it being safer. (I spend a lot of time trying to educate users here - and it make me happy to have an new user with only 8 posts ask a very good question!!)

 

You might not realize, and I didn't explain in detail, but the process I laid out is recoverable.

 

You don't have to do this, but it is easier if BEFORE THE NEW CONFIG, you stop the array and make a backup of your config folder on the flash disk (array must be stopped when the backup is taken).

 

If a drive failed during the parity build, you could restore your config folder and the array would be back as it was, with the 5TB drive as parity and the 3TB failing disk back in the array. The failed disk might not even appear failed. But you could swap it out and rebuild it as best you could, if the 3T drive held out.

 

The one rule is that you would not be able to do writes to the array while the parity is rebuilding. Including not formatting the 5TB parity drive until parity is built (I did mention that).

 

If you do not do the writes, reverting back would be successful and you'd be able to recover a failed disk.

 

Sometimes going into these details creates more confusion and raises more fears than just giving the instructions. I've been dong this a long time and never had a disk fail under these circumstances.  Much more common is a disk would drop offline due to a cabling problem. (That can be fixed and the parity build restated.)

 

I am not 100% sure the recoverability if there is a failure in the parity swap. I guess in theory it would maintain both parities until the parity is rebuilt. But what if another drive failed during the parity build. And then the marginal 3T drive failed trying to rebuild it. The 3T data is on the DVR, but it won't help with the second drive's data.

 

Generally I don't like parity swap with a failing disk because you are building parity using the failing drive. How much longer is it going to keep working? If it fails in the middle of the parity rebuild the array would stop and I don't think unRAID would leave you in a useful state. Backing up the drive to the DVR first protects you, but a lot of people in your shoes don't have an extra drive to do that. My process puts FAR less stress on the failing drive. The only thing we do with it is read its data - once.

 

The other thing I like about copying its data rather than rebuilding it - is that is only a portion of the data is copied over, you have that data safe. You can choose the most precious parts first. And I've had recovery situations where the disk would get warm and stop working, and I'd have to take it offline, even put it in the fridge, to get it working again. On a rebuild, unRAID is rebuilding the disk sector by sector, and you're never really sure what files are intact until it is completely finished.

[trurl]

41 minutes ago, bjp999 said:

I don't like parity swap with a failing disk because you are building parity using the failing drive

Parity swap doesn't build parity, it copies it from the original parity.

[Griff1324]

What is a noob to do?   I have two highly active members of these forums directing me in two different directions!  I appreciate all of the help and feedback guys.

[SSD]

Even the experts disagree sometimes. I have never used the parity swap process - but even if it does indeed copy parity, it does not really change much of my argument. But if you are backing up the failing 3T disk before you do anything, I'd say risk is similar. I'm just not much on rebuilding data disks. I prefer to let data get copied to them the old fashioned way and rebuild parity. But either way will get you what you want. I don't think there is a wrong choice in your case.