Hello Lime-Tech community, this is my first post here.  I stumbled on to unRaid a couple months back thanks to LinusTechTips and I've been very happy so far.  The question I have for you all is, what is the best way to upgrade my storage drives?

 

Presently I have 3x3tb WD reds, all running without parity drive using a 240gb flash drive for cache.

 

I have 4x HGST 6tb drives that I plan to run together, 3 in the raid 1 as parity with the same cache drive.

 

I presently have about 6TB of data in various folders on the existing array.  Would it make sense to first add a 6tb drive for parity, and then 1 by one swap out the existing 3tb drives or hook up a system and pull the files off via cable network (I'm wireless right now) onto one of the 6tb drives, install 3 as a fresh array, and then put them back on the same way and add the final 6tb drive for parity as the final step?

 

thanks in advance!

Hello and welcome.  The fastest way would probably be to add the 6TB disks to the array and then do a disk to disk copy from the 3TB disks to the 6TB disks.  When that is complete you'd do a New Config and add parity.  If that's not practical due to space or connections then adding parity now and rebuilding onto the 6TB disks is next best.  Copying over the network from another machine is also possible.

Is there a function in unraid that allows disk copy?  Or is this something that is commandline driven?  My license only allows 5 disks I believe, which could be the other issue.

You can use Midnight Commander (mc from the command line, with a utility like screen if you want), and there are Dockers like Krusader that can help - both are 2 panel file manager type applications.  If you have a Basic license, it supports up to 6 attached storage devices.  If you are up against a device limit, the rebuild or copy over the network strategy will be best.

You can add 2 6TB to your current 4 drives and be under the license limit. 3x3=9, 2x6=12, so all your data will fit on those two new drives. Then you can do a "new config" pull the 3 old drives, add the other two new drives, assign 1 as parity and let it build.

 

I would use the command line with rsync to copy the data, with the correct switches it will verify the copy with checksums.

Do you recommend preclearing each disk as I install them?

21 minutes ago, Type R said:

Do you recommend preclearing each disk as I install them?

It definitely doesn't hurt to do a bit of shake-down before you put your trust in a new drive.

19 hours ago, jonathanm said:

You can add 2 6TB to your current 4 drives and be under the license limit. 3x3=9, 2x6=12, so all your data will fit on those two new drives. Then you can do a "new config" pull the 3 old drives, add the other two new drives, assign 1 as parity and let it build.

 

I would use the command line with rsync to copy the data, with the correct switches it will verify the copy with checksums.

 

Would you have instructions for this somewhere?  I think this is the way I'll go. 

1 hour ago, Type R said:

 

Would you have instructions for this somewhere?  I think this is the way I'll go. 

I use a 2 part process, first command copies, the second command verifies but doesn't copy anything, it just lists what doesn't match, and it's up to you to figure out why it didn't match, and either copy again or ignore it.

 

If you are less paranoid than I, you can change the rsync switches to do the checksum on the original copy, and trust that everything worked. I like the 2 pass method, it gives me the warm'n'fuzzies to see an empty verify text file indicating everything is identical from source to destination since there was nothing changed. It does, however, take twice as long.

 

As I outlined in that series of posts, all the output is sent to a text file so you can browse through and search the list of things copied.

 

In your situation, you could easily send the contents of 2 of the drives to 1 of the new drives, and then the last drive will just be 1 to 1. Or, if you want to organize things differently, you just need to parse the folder structure and copy only the folders you want to specific drives. Just add the folder names to both the source and destination parts of the command, being careful to properly use a closing slash or not on the path.

If you are not in a big hurry, you can install the 6TB drive as parity and let it build parity.  After that is done, shutdown the server and remove   *one* of the 3TB drives (set it aside as a backup until you are completely finished!) and replace it with a 6TB drive and it will rebuild the data on it.  Repeat until you are finished.  (NOTE:  double check all SATA connectors after each disk change if you are not using hot plug cages!

 

It will probably take 12 to 14 hours for each rebuilt so you only need to be there for the first two minutes. 

 

Advantages-- This is virtually foolproof for anyone and very simple to do and you will have a backup of your data when you are done. After about a week, run a parity check on the new array to make sure that you don't have any issues.  (You should have zero errors.) 

 

After you are satisfied that everything is stable, you can do what you want with those old 3TB drives. 

unOk, I think I blew this somehow.  I precleared the first drive, and I stupidly selected and added it to the array.  I am now trying to figure out how to unassign it without wrecking my existing array.  I don't have a parity drive, and I have now used up all of my PSU outlets, so can't install another drive.

 

Is there a way to uninstall this drive from the array without effecting anything?

Why do you want to remove it from the array? I thought you wanted to copy the data from 2 of your 3TB drives to that new 6TB that you just added?

 

With no parity installed, you can just do a new config to assign drives to whatever data slots you want.

31 minutes ago, jonathanm said:

Why do you want to remove it from the array? I thought you wanted to copy the data from 2 of your 3TB drives to that new 6TB that you just added?

 

With no parity installed, you can just do a new config to assign drives to whatever data slots you want.

 

In doing that, do I not lose the data on those drives?

8 minutes ago, Type R said:

 

In doing that, do I not lose the data on those drives?

Doing what? New config won't erase any data unless you assign a disk to the parity slots, if you assign it to parity it will be erased as parity is built to it.

 

Adding a disk to an array with parity disks already defined will zero the disks so parity remains valid when the disk is added.

ok, thanks for clarifying, my thought was in defining a new config, the old one would be lost

15 minutes ago, Type R said:

ok, thanks for clarifying, my thought was in defining a new config, the old one would be lost

Yep. The config is the assignment of which physical disk is assigned to which slot number. A new config invalidates the old slot numbering and allows you to assign slots as you see fit. No data is altered on unraid format compatible disks unless you assign them to a parity slot.

 

The old config listing slot numbers and the corresponding physical drives is lost.

So I removed the 6TB drive from the array, and then added back as a parity drive.  Upon adding it, I ran a check, and there are over 67M errors however the system claims the parity is valid?  I looked at the log, and the only thing of relevance is:

 

Jan 9 00:18:42 ArcaMax kernel: md: recovery thread: P corrected, sector=664
Jan 9 00:18:42 ArcaMax kernel: md: recovery thread: P corrected, sector=672
Jan 9 00:18:42 ArcaMax kernel: md: recovery thread: P corrected, sector=680
Jan 9 00:18:42 ArcaMax kernel: md: recovery thread: P corrected, sector=688
Jan 9 00:18:42 ArcaMax kernel: md: recovery thread: P corrected, sector=696
Jan 9 00:18:42 ArcaMax kernel: md: recovery thread: P corrected, sector=704
Jan 9 00:18:42 ArcaMax kernel: md: recovery thread: P corrected, sector=712
Jan 9 00:18:42 ArcaMax kernel: md: recovery thread: P corrected, sector=720
Jan 9 00:18:42 ArcaMax kernel: md: recovery thread: P corrected, sector=728
Jan 9 00:18:42 ArcaMax kernel: md: recovery thread: P corrected, sector=736
Jan 9 00:18:42 ArcaMax kernel: md: recovery thread: P corrected, sector=744
Jan 9 00:18:42 ArcaMax kernel: md: recovery thread: P corrected, sector=752
Jan 9 00:18:42 ArcaMax kernel: md: recovery thread: P corrected, sector=760
Jan 9 00:18:42 ArcaMax kernel: md: recovery thread: P corrected, sector=768
Jan 9 00:18:42 ArcaMax kernel: md: recovery thread: P corrected, sector=776
Jan 9 00:18:42 ArcaMax kernel: md: recovery thread: P corrected, sector=784
Jan 9 00:18:42 ArcaMax kernel: md: recovery thread: P corrected, sector=792
Jan 9 00:18:42 ArcaMax kernel: md: recovery thread: P corrected, sector=800
Jan 9 00:18:42 ArcaMax kernel: md: recovery thread: P corrected, sector=808
Jan 9 00:18:42 ArcaMax kernel: md: recovery thread: P corrected, sector=816
Jan 9 00:18:42 ArcaMax kernel: md: recovery thread: P corrected, sector=824

 

Now if I'm reading this correctly, its telling me there are bad sectors in the parity drive, which I don't understand, because I just finished preclearing it and didn't have any issues.

5 hours ago, Type R said:

Now if I'm reading this correctly, its telling me there are bad sectors in the parity drive

No, it's telling you there were parity sync errors that were corrected.