Hello all,

I'm hoping I can get some quick guidance on some maintenance I'm about to perform. Really I guess it's what's advised as to what steps in what order.

Quick rundown:

6 x 12TB disks = 2 parity (P) + 4 data (D)

6 drive slots in server chassis

Disks are SAS so cannot use USB SATA adapter to temp add an unassigned disk (dang it)

P1 - Dead - red mark in MAIN, no RWs

P2 - Working

D1 - Dead - red mark in MAIN, no RWs - any requested RW comes from P2. ALL data now 'moved' to D3 & D4

D2 - Dying - lots of read errors, but still operational. MOST data 'moved' to D3 & D4

D3 - Working

D4 - Working

I have 2 x 10TB ready to replace.

Where I want to end up:

P1 - Gone

P2 - But now P1 (12TB)

D1 - Replaced with new 10TB

D2 - Replaced with new 10TB

D3 - Unchanged

D4 - Unchanged

However, due to having no spare slots, I need to get 1 x 10TB in now so I can fully drain D2 before it dies a death on me. So, I think what I need to do is:

Step 1

Remove dead 12TB D1 from array

Step 2

Add new 10TB D1 to array

Step 3

Allow parity to rebuild if it needs to

Step 4

Drain remaining data (~3TB) from D2 to D1

Step 5

Remove dead 12TB D2 from array

Step 6

Add new 10TB D2 to array

Step 7

Do some rebalancing with Unbalance

Step 8

Remove P1 from array - does this mean parity will need to rebuild again, as it'll now be single parity - do I need to make config changes somewhere for that to happen?

Step 9

Change P2 to P1 - is this just a name change? Will parity need to be rebuilt for a 3rd time?

Step 10

Am I done???

I hope this all makes sense. I know what I'm trying to do is easy, but I'm super twitchy as I've not done anything like this for about 5 years since I built the original server, and because I'm mucking about with parity AND data disks. I REALLY appreciate any help the lovely community can give me!

Thanks!

Edited Thursday at 01:59 PM4 days by TheBenga

Your intended final setup is possible, but I would change the order.

First, a 10 TB disk cannot directly replace a 12 TB disk through a rebuild, even after all files are moved off. Using the smaller disks requires copying the files and then reconfiguring the array.

Also, changing Parity 2 to Parity 1 requires a new parity sync. They use different calculations, so it is not just a name change.

The immediate priority is the remaining data on Disk 2. With P1 and D1 disabled, another unreadable disk can prevent recovery. D1’s emulation depends on P2 and the other data disks, including D2.

You may already have space for a temporary destination: the physical bay occupied by disabled P1. Subject to confirming the disk identities and connection compatibility, one 10 TB disk could go there and remain outside the array, mounted through Unassigned Devices. That would let you copy the remaining D2 files without first replacing D1 or rebuilding parity.

Before changing anything, please attach Tools → Diagnostics and a screenshot of Main. A red mark means disabled; it does not by itself prove that the drive has physically failed. Please also confirm whether the earlier moves completed without errors and whether you have an independent backup.

For now, avoid New Config, parity rebuilds, and rebalancing. The priority is to copy the remaining files, retain the originals, and verify the copies.

Once the required files are safe, we can plan the final assignments and a single final parity sync. A disk containing recovered files needs careful handling: adding it normally to a parity-protected array would clear it.

Awesome advice. I must admit I didn't think of adding a disk into the main chassis but leaving as unassigned.

Here's the pic and diags you requested.

Thank you very much indeed for your input.

bakery-ur-diagnostics-20260924-1523.zip

Sorry, you asked some other Q's too.

No independent backups. But there is now very little on D2 that isn't recoverable from public sources. I've already moved all personal and private data. Obvs I'd rather not have to recover what's left so getting the remaining data off D2 is defo priority. FYI, not all the used data on D2 is wanted, I've just done a quick check and it's about 4TB to come off it, the other 600Gb can be dropped.

The bad disks are defo dead. They can no longer run a SMART test and are quite long in the tooth re: runtime (~68,000h each)

Thanks again!

In addition to parity and disk1, disk2 does also appear to be failing. Its SMART report warns of impending failure, records a lot of uncorrected read errors, and shows two failed short self-tests. I would prioritize copying the remaining wanted files rather than running more tests.

Stop the array and shut down, then replace the disabled P1 disk physically with one of the 10 TB disks. Leave the new disk unassigned, with the other array assignments unchanged. Prepare only that new blank disk through Unassigned Devices, preferably as XFS, then restart the existing array and copy the wanted files from disk2 to it.

Use copy rather than move, keeping the originals for now. If some files produce read errors, record which ones failed and continue with the readable files. With P1 and D1 already disabled, parity cannot reliably recover unreadable sectors on D2.

One other detail: the diagnostics show about 878 GiB still used on emulated disk1. Please check whether that contains anything you still need, or whether everything there already has a good copy elsewhere.

For now, avoid New Config, parity checks/rebuilds, and rebalancing. Once the wanted files are safe, we can arrange the final configuration and build parity once. Do not add the populated transfer disk through the normal add-disk procedure, because that would clear it.

Since there is no independent backup, I would also make a separate copy of the personal files now on D3/D4.

Grand, and yes, agree getting the data of the dying D2 is a priority.

Confirmed what's left on D1 is junk, that disk was drained quite some time ago.

Back in a bit!

All steps so far good, going to kick off the final drain of D2 -> UD1

n00b question - it makes sense that Unbalance plugin cannot see UD1 as it is not in the array. I have restarted the Unbalance server in case it picks it up since I formatted and mounted UD1 but it doesn't see it.

I was hoping I could use it for logging, perm checks and ease of selection purposes. But if that's not possible, would you suggest using MC from terminal as the next best approach?

Cheers,

MC would be my choice; but you can also use the built-in file manager.

3 hours ago, TheBenga said:

All steps so far good, going to kick off the final drain of D2 -> UD1

n00b question - it makes sense that Unbalance plugin cannot see UD1 as it is not in the array. I have restarted the Unbalance server in case it picks it up since I formatted and mounted UD1 but it doesn't see it.

I was hoping I could use it for logging, perm checks and ease of selection purposes. But if that's not possible, would you suggest using MC from terminal as the next best approach?

Cheers,

I personally would use the tmux plugin for a persistent terminal and then use rsync -aHAXx --numeric-ids --info=progress2 /mnt/disk2/ /mnt/disks/UD_MOUNTPOINT/ as this will copy the data over.

Edited Thursday at 07:23 PM4 days by MowMdown

Back!

So, all copies completed.

Next:

1. Remove 12TB D2 from array

2. Add new unused 10TB disk to array

3. Run parity build

4. Copy UD1 data to array

?

Happy to be wrong!

Now that the copies are complete and nothing wanted remains on D1, I would use Tools → New Config and include the UD disk directly in the array. That avoids copying its files back again.

The intended assignments would be:

- Parity: the working 12 TB disk currently assigned as P2.

- Disk 1: the 10 TB UD disk containing the recovered files.

- Disk 2: the unused 10 TB disk.

- Disk 3 and Disk 4: unchanged.

The old P1, D1, and D2 would be left out.

Unmount UD1 from Unassigned Devices before assigning it through New Config. Leave “Parity is already valid” unchecked, because this layout requires a fresh parity sync.

Only the unused 10 TB disk should need formatting. Do not format UD1, D3, or D4. If any populated disk appears unmountable, stop and post new diags.

Make sure to include UD1 during New Config, before starting the new array. Adding it afterward through the normal add-disk process would clear it. Keep the removed disks unchanged until you have verified the recovered files.

Makes sense.

Can I just clarify, at what stage in the above should I be physically removing D1?

Should I be now doing:

1. Shut down the array and power off

2. Remove current D1 - this disk we've made no changes, access or other actions so far, it died a while back and I 'moved' the data before this thread started (albeit the rw's actually came from parity as D1 was dead)

3. Add new 10TB into now free slot that D1 was in

4. Power up (with array auto-off - I always set it like this anyway)

5. Format new 10TB we just introduced

6. Use new config to set as above:

7. - Parity: the working 12 TB disk currently assigned as P2.

   - Disk 1: the 10 TB UD disk containing the recovered files.

   - Disk 2: the unused 10 TB disk.

   - Disk 3 and Disk 4: unchanged.

8. The above is the first time both new disks are being added to array config, and we're adding them at the same time, in one go

9. Ensure "parity is valid" is unchecked

10. Start the array

11. Wait for parity!

Sorry to be so granular, just being super careful 😊

Thanks,

5 hours ago, TheBenga said:

(albeit the rw's actually came from parity as D1 was dead)

actually came from the parity calculation by reading all other array drives including parity. Parity by itself has none of your data.

5 hours ago, TheBenga said:

Format new 10TB we just introduced

While you could technically format the 10TB as an Unassigned Device, might as well wait until after it is assigned during New Config.

Assign new 10TB when doing New Config, then format it after starting the array. You can do the format during or after parity build.

11 hours ago, JorgeB said:

Only the unused 10 TB disk should need formatting. Do not format UD1, D3, or D4. If any populated disk appears unmountable, stop and post new diags.

Hi guys,

Soooo, all good so far (he says very tentatively!)

Disks are in as we agreed, parity is rebuilding (usually about 17h). I've checked a pile of files and Got a question on this step:

--Assign new 10TB when doing New Config, then format it after starting the array. You can do the format during or after parity build

The new 10TB disk is in (assigned to D2), and as expected is showing as unformatted. How do I format it now during the parity build as suggested?

Cheers!

Click the "format" button, next to array start/stop, make sure it's the only disk listed as the one to be formatted.

omg, so sorry, I was looking in the disk settings. Got it, and done.

The array is fully back online, parity complete, lights are green, 5-5-5.

Thank you both so so much!