OK, this looks bad. What do I do next?

[JorgeB]

You can start the array and copy files to it, just copy to any of the other disks, not disk5 (you'd need to format it first anyway).

 

I have some doubts if your parity2 was 100% in sync, so after the copy (and if you can recover most data) probably best to do a new config and a new parity sync.

[Karyudo]

OK, next question: How do I look at what's on (flaky) Disk 5, and how do I copy stuff to one of the other disks explicitly? I only see user shares; there are no exportable disk shares (and never have been).

[JorgeB]

On the UD plugin there's an export toggle, enable it and the disk will appear under shares.

 

You can also enable disk shares, just be aware that you should never copy from a disk share to a user share, you can copy from the UD share to a disk share.

[Karyudo]

Just to be clear, it's OK to copy from the 'disk5' share created under UD to a user share, correct?

 

Next annoying problem: How do I get past the "Network Credentials" login page that pops up in Windows 10 when I click on the disk5 share? My usual username and password doesn't work; neither does root/<none>. Or any number of other combinations. What's this damn thing looking for??

 

 

[JorgeB]

You can always copy from a UD share to an array disk or user share, problem is between disk and user shares inside the array.

 

As for the credentials, go to settings -> unassigned device settings and check "smb security", if enable disable it and click apply, if disable enable, apply, disable, apply, it should work after that.

[Karyudo]

Is there a place to set SMB security for everything? Because now I can't log in to *any* shares (even all the array shares I previously could).

 

I can't remember the last time I was as frustrated about something as I am with this nightmare. Once again, thank you for your patience.

[JorgeB]

I'm not close to a sever and don't remember exactly where they are, but unless you changed something it's more likely Windows trying to use the wrong credentials, go to control panel -> credential manager and delete all server credentials then reboot the desktop and try again.

[Karyudo]

Between my last post and yours, I restarted my Windows machine. Navigated to a share, entered *the exact same* credentials one more time, et voila: success. Makes no sense to me, but whatever.

 

I just tried copying something from disk5 to the array's Videos share, and it already exists. I tried playing the file from the array, and it works. Why is that the case? Why aren't the files on disk5 unique? I thought all files on disk5 would be missing completely from the array. Can you explain what's on disk5 vs. what's on the array, and how best to make sure I keep the working copy of each file (without having to manually sift through 2TB of duplicates)?

 

[JorgeB]

If you are really copying from the unassigned disk5 and the array disk5 is still unmountable duplicates can't exist, unless they were there before, in doubt create a new folder or share and copy there, then sort it out later when the array is restored.

[Karyudo]

Copying now to the 'restored' directory created during the failed recovery attempt this morning. Looks like it's going to take about 11 hours. See you tomorrow morning, I guess!

[Karyudo]

Copying completed: just one problem file out of 2TB. Not sure which file; don't care that much, since it turns out they were all recently downloaded.

 

Looking at the Main tab, though, it sure looks as though everything I've just copied has turned up on the old Disk 5! HOW IS THAT POSSIBLE?!

 

 

I guess that has to be fixed first. But then what's next? Move to XFS, then redo parity? Or redo parity and then move to XFS?

[JorgeB]

Post a screenshot of currant array so it's easier to see current state.

[Karyudo]

Oops! Meant to; had it all ready to go, and then didn't attach it.

[JorgeB]

So disk5 is monted after all, I though it was still unmountable, looks like you copied the info to the emulated disk.

 

Also it's a little odd that both the emulated disk5 and the old disk5 have the exact same used space.

 

Are the other 2 unassigned disks parity and disk5 replacements?

[Karyudo]

This doesn't make any sense at all to me. I didn't do anything dumb when copying—I followed your instructions exactly. The 'restored' directory we created was created without any Disk 5 in place! And I mounted the old Disk 5 in Unassigned Devices, not as part of the array! What part of this doesn't unRAID understand?!? If anything screwed up here, it's unRAID.

 

Yes, the other two unassigned disks are the unsuccessful disk5 rebuild (-LCHC) and a parity replacement (-MZHK).

 

[JorgeB]

I've noticed some weird BTRFS behavior before, eg, when re-assigning a btrfs cache disk to the array, so just to make sure that the data is really on the array, shutdown the server, disconnect old disk 5 and power back up again.

[Karyudo]

Shut down. Disconnected (SATA, power) old Disk 5. Restarted array. The 'restored' directory is there, but not all of the sub-directories or all of the files. Some are, but some aren't. Another screenshot is attached.

 

I'm satisfied that I can get almost all of the files off the old Disk 5 when I want to, but I guess the BTRFS weirdness should be solved first? Unless there's a way to force a copy onto other-than-old-Disk-5 array space?

[JorgeB]

This is what I would do:

 

-stop array

-tools -> new config (select keep current assignments)

-reassign new parity and new disk5, double check all assignments.

-click on disk5

-change filesystem to xfs

-start array to begin parity sync

-format disk5 (make sure it's the only disk listed next to the format button)

 

When sync is done and the array is protected reconnect old disk5 with the UD plugin and copy all the data to the array, since new disk 5 is now xfs there shouldn't be any more issues.

 

Then if you want you can proceed with converting the other disks to xfs, disk6 is easy because it's almost empty, the others will take more time.

 

Although I don't recommend btrfs for data disks in your case I think that parity was corrupted when the disks/controller crashed, it's a shame you don't have the syslog, because it's the only logical explanation for the rebuilt disk having all 3 superblocks failing crc and making recovery impossible.

[Karyudo]

Yeah, maybe if I'da not done or cancelled the parity update earlier, I'd have been in better shape. Seems like kind of a failure of unRAID that the syslog didn't show the activity of interest.

 

I'll get started on the list of Things To Do that you've outlined—thanks!

[Karyudo]

Parity rebuild started. (Note: When I did the New Config, there was no explicit, simple "Keep current assignments" option. In the sub-menu of checkboxes for "Keep current assignments," I selected "All.")

 

New (-LCHC) Disk 5 installed, formatted XFS.

 

But what's the orange triangle beside Disk 5 doing there, saying "Device contents emulated"? I thought the point of formatting (especially formatting XFS rather than BTRFS) was to blow away whatever Disk 5 history there was?

 

[JorgeB]

You didn't click apply on new config, instead it's rebuilding disk5 again (and syncing parity).

 

Since parity2 can't be trusted you can cancel and do the new config, but if you prefer you can leave it like that, parity1 will be good, if there are errors on parity2 they will be corrected after the first parity check.

[Karyudo]

Oh, so unRAID doesn't work like every other program ever, where clicking 'OK' also applies the changes? You actually gotta explicitly click on 'Apply' before exiting? That's lame.

 

I think I'm going to cancel and start again. Thanks for the analysis!

 

[Karyudo]

Interesting. Figured something out:

 

You know why I didn't click on 'Apply'? Because it was greyed out. Do you know why it was greyed out? Because I hadn't clicked on the checkbox for 'Yes, I want to do this.' Do you know why I didn't click on 'Yes, I want to do this'? Because I didn't see it. Do you know why I didn't see it? Because it was covered up by the secondary checkbox for 'Keep existing configuration.' Do you know why it was covered up by the secondary checkbox? Because I didn't click on '...close' at the bottom of the checkbox. Do you know why I didn't click on the '...close' at the bottom of the checkbox? Because I didn't know I needed to. Do you know why I didn't know I needed to? Because the interface design for this task is poor.

 

It's pretty dumb to cover up an important part of the UI ("Yes, I want to do this") with what looks like a wholly optional step ("...close"). Especially when there's an ambiguously-titled button ("Done") to click that doesn't then actually apply changes.

[Karyudo]

OK, this looks better, I think. Both parity drives rebuilding, Disk 5 with a nice green dot.

 

I didn't get a chance to re-format Disk 5 during this cycle. Clicking on 'View' shows nothing on the drive, though. I imagine by canceling the original parity check and rebuild, that leaves Disk 5 back where it was immediately following the XFS format step?

[JorgeB]

Yes, looking normal.