[SOLVED] Problem creating cache drive

[smzmtyn]

Hi folks,

 

i just installed UNRAID in trial mode. So far I've been able to add my 2 main disks (2TB xfs encrypted) and a 3rd disk as a parity disk (3TB). This seems to be fine. I'm having issues trying to set up a cache drive however:

 

What I did:

I bring down the array, then assign a 240 GB SSD to the cache pool, and then start the array.

Result:

Cache drive appears in pool and says "Unmountable: No file system" and FS is set to auto

 

Ok, no problem, I'll just format it..

 

What I did:

Under "Format will create a file system in all Unmountable disks.", I tick "Yes, I want to do this" and proceed with the format.

Result:

Cache drive says "Formatting" for a few seconds, followed again by "Unmountable: No file system".

 

Ok, maybe auto is not the right fs, so let me meddle with that...

 

What I did:

Click on the cache drive, set File system type to xfs - encrypted, then hit Apply and Done, then on the Main screen, run Format again.

Result:

Cache drive says "Formatting" for a few seconds, followed now by "Unmountable: Missing encryption key".

 

Ok, hmm I needed to provide a key before, maybe I just need to do that again?

 

What I did:

Stop the array, provide the encryption passphrase, then start the array.

Result:

Cache drive says "Unmountable: Volume not encrypted"

 

I tried formatting the cache drive again after the message above, but it goes back to "Unmountable: Missing encryption key". Btw, I get stuck in the same loop if I use xfs encrypted for the cache fs also. I'm at my wits' end, would appreciate some help here.

Edited December 6, 2021 by smzmtyn

[trurl]

attach diagnostics to your NEXT post in this thread

[Squid]

Post your diagnostics

[smzmtyn]

Appreciate the super fast replies. Diagnostics file attached.

 

Edited December 4, 2021 by smzmtyn

[trurl]

23 minutes ago, smzmtyn said:

I've been able to add my 2 main disks (2TB xfs encrypted) and a 3rd disk as a parity disk (3TB). This seems to be fine

Maybe you didn't notice since you didn't mention, but it isn't fine.

 

You were having (connection?) problems on multiple disks the several times you tried to build parity. You should be able to see these in the Errors column on Main.

 

Looks like you are using old disks. Nothing necessarily wrong with that, but both of your data disks have pending sectors. You should be able to see warnings for those on the Dashboard page.

 

Your SSD is perhaps the oldest of all, it even says it has no life left. Perhaps that is the reason for this

Dec  4 06:37:17 svr root: ERROR: '/dev/sdc1' is too small to make a usable filesystem
Dec  4 06:37:17 svr root: ERROR: minimum size for each btrfs device is 47185920

and your problems getting cache to work.

 

Also noticed a couple of times where you tried to assign a flash drive to array or cache. Don't do that.

 

I suggest starting over, and if you want to use those spinners, preclear them to make sure they pass and to perhaps get those pending sectors realloccated. And use a different SSD.

 

 

[smzmtyn]

4 minutes ago, trurl said:

Maybe you didn't notice since you didn't mention, but it isn't fine.

 

You were having (connection?) problems on multiple disks the several times you tried to build parity. You should be able to see these in the Errors column on Main.

 

Looks like you are using old disks. Nothing necessarily wrong with that, but both of your data disks have pending sectors. You should be able to see warnings for those on the Dashboard page.

 

Your SSD is perhaps the oldest of all, it even says it has no life left. Perhaps that is the reason for this

Dec  4 06:37:17 svr root: ERROR: '/dev/sdc1' is too small to make a usable filesystem
Dec  4 06:37:17 svr root: ERROR: minimum size for each btrfs device is 47185920

and your problems getting cache to work.

 

Also noticed a couple of times where you tried to assign a flash drive to array or cache. Don't do that.

 

I suggest starting over, and if you want to use those spinners, preclear them to make sure they pass and to perhaps get those pending sectors realloccated. And use a different SSD.

 

 

 

Appreciate the input here.

 

Apologies, yes, I had issues with the spinners initially, but while I do see some errors in their errors columns, the SMART status is reporting as ok, so I thought all was well. This SSD was used in a Proxmox system as both the main boot device as well as VM storage, so it seems like it should be fine, but will try another. I'm using the Preclear plugin to pre-clear all the disks now, fingers crossed.

 

Thank you for the assist!

[trurl]

5 minutes ago, smzmtyn said:

SMART status is reporting as ok

Where are you looking that says they are OK?

23 minutes ago, trurl said:

You should be able to see warnings for those on the Dashboard page.

 

[smzmtyn]

10 minutes ago, trurl said:

Where are you looking that says they are OK?

 

I think after initially having the issues, at some point, I clicked into the drives, and though there were errors, the overall SMART status reported ok. You're right though, if I go to the Dashboard page, they all show as having errors. Perhaps I'm putting too much stock into overall SMART health.

[trurl]

Each data disk in Unraid is a separate filesystem that can be read independently of all others.

 

But...

 

Parity (in any system) is just an extra bit that allows a missing bit to be calculated from all the other bits. To reliably rebuild a disk, every bit of parity plus every bit of all other disks must be reliably read. So, all disks in your array need to be reliable.

 

Unraid monitors certain critical SMART attributes and these are what are shown as warnings on the Dashboard page.

 

You should setup Notifications to alert you immediately by email or other agent as soon as a problem is discovered. And take care of problems as they occur. Don't let one problem become multiple problems and data loss.

 

[trurl]

I also noticed you had already enabled Docker and VM Manager in Settings. Disable these until you get everything else working well.

 

If you enable these without cache their shares (appdata, domains, system) will be created on the array, which will impact docker/VM performance and keep array disks spunup since these files are always open.

[smzmtyn]

2 hours ago, trurl said:

Each data disk in Unraid is a separate filesystem that can be read independently of all others.

 

But...

 

Parity (in any system) is just an extra bit that allows a missing bit to be calculated from all the other bits. To reliably rebuild a disk, every bit of parity plus every bit of all other disks must be reliably read. So, all disks in your array need to be reliable.

 

Unraid monitors certain critical SMART attributes and these are what are shown as warnings on the Dashboard page.

 

You should setup Notifications to alert you immediately by email or other agent as soon as a problem is discovered. And take care of problems as they occur. Don't let one problem become multiple problems and data loss.

 

Really appreciate the insight here. Per what you said, I went ahead and ordered a couple more drives for both cache and main data. Follow up question here - I was able to pre-clear the 240 GB SSD successfully (no errors in the log), does that imply anything about the health of the drive, does that mean it's safe to use?

 

Thanks also for the suggestion on disabling Docker & VMs, will keep that in mind for the future.

[trurl]

3 hours ago, smzmtyn said:

pre-clear the 240 GB SSD

Sorry forgot to mention preclear not usually done for SSDs but it might have been good in this case.

 

Earlier it seemed that it didn't have the capacity to become a cache drive. Now that I think about it, and based on how you said it had been used, probably it had multiple partitions on it. Preclear should have fixed that though.

 

It is pretty old, but it didn't claim to be failing. I guess you could try it since it won't be in the array where it can compromise the ability to rebuild other disks. Depends on how you intend to use cache I guess. Unless you have a multidisk pool for that it won't have any redundancy.

[smzmtyn]

On 12/4/2021 at 6:59 PM, trurl said:

Sorry forgot to mention preclear not usually done for SSDs but it might have been good in this case.

 

Earlier it seemed that it didn't have the capacity to become a cache drive. Now that I think about it, and based on how you said it had been used, probably it had multiple partitions on it. Preclear should have fixed that though.

 

It is pretty old, but it didn't claim to be failing. I guess you could try it since it won't be in the array where it can compromise the ability to rebuild other disks. Depends on how you intend to use cache I guess. Unless you have a multidisk pool for that it won't have any redundancy.

Ahh dang, I'll have to see if I can cancel the order for the cache drive I got. Thanks though.

 

PS, I was finally just able to resolve this... When I dropped into a terminal and ran `lsblk` on the SSD, it oddly still had all the Proxmox partitions on it, despite the preclearing and my prior attempts to format it. I then used `fdisk` to delete the existing partitions and create a new one, and restarted. Upon restarting, unRAID said something about the partition being incorrect/unusable, but I was then successfully able to use the built-in formatting utility (on the Main page) and use the drive as a cache drive. Interestingly, the formatting step this time actually took a bit of time (few mins) vs with my prior attempts where it was less than 30 seconds (likely because it was erroring almost immediately). Hope it helps someone in a similar situation.