Replace one cache SSD with two new SSD's.

[YourNightmar3]

A couple days ago i made a post where it turns out there is something wrong with one of my cache SSD's: 

I haven't fully figured out the problem yet but i purchased two new SSD's because i have been wanting to replace my cache drives for a while anyway. Then i'll figure out in the other thread why one of the two current SSD's isnt working.

 

My current situation: One working cache SSD and two new unassigned SSD's all connected to the server. How do i replace the one cache disk with the two new ones?

 

Can i add the two new SSD's to the cache pool, wait for the BTRFS action to complete, then remove the current cache SSD from the pool so i only have the two new ones left? I'd like to avoid transferring all cache data to my array if possible.

 

Thanks.

 

Edited December 12, 2022 by YourNightmar3

[JorgeB]

11 hours ago, YourNightmar3 said:

Can i add the two new SSD's to the cache pool, wait for the BTRFS action to complete, then remove the current cache SSD from the pool so i only have the two new ones left? I'd like to avoid transferring all cache data to my array if possible.

Yes, assuming the current one is using btrfs.

[YourNightmar3]

47 minutes ago, JorgeB said:

Yes, assuming the current one is using btrfs.

Yes it is, thanks a lot. Really couldnt find this specific situation by googling and i just needed a confirmation before doing it

[YourNightmar3]

Hmm i just started the BTRFS action and now after about 30 minutes my server is entirely unreachable on the network. My KVM shows no video signal but the machine is still running. I assume a shutdown during the BTRFS action is unsafe? Or can do a hard shutdown wioth the power button and restart the server? I did watch it start and it was running for 10+ minutes.

Edited December 13, 2022 by YourNightmar3

[JorgeB]

35 minutes ago, YourNightmar3 said:

I assume a shutdown during the BTRFS action is unsafe?

Not ideal but wait a few more minutes and if still nothing it's the only option, I understand there's no local console access also?

[YourNightmar3]

19 minutes ago, JorgeB said:

Not ideal but wait a few more minutes and if still nothing it's the only option, I understand there's no local console access also?

 

Yeah no console access unfortunately. I have rebooted the server and a parity check has just started. I don't see anything about a BTRFS operation, how do i know if it ran successfully? I do see my cache pool with my two new drives and one old SSD, all three of those devices say "normal operation, device is active". But i'm also getting a "Warning - pool BTRFS too many profiles (You can ignore this warning when a pool balance operation is in progress)" message while there's no new BTRFS action starting at the moment, just the parity check.

 

How do i know if the previous BTRFS operation finished or not? Do i need to somehow start a balance operation manually (seeing this warning)? And do i need to wait for the parity check to finish first?

[JorgeB]

Post current diags.

[YourNightmar3]

8 minutes ago, JorgeB said:

Post current diags.

Ok here it is tower-diagnostics-20221213-1147.zip

[JorgeB]

Balance was resumed and is still going at this moment, lets see if it finishes this time.

[YourNightmar3]

1 hour ago, JorgeB said:

Balance was resumed and is still going at this moment, lets see if it finishes this time.

 

I ran this command earlier:
 

btrfs balance status /mnt/cache

 

And it said 9% left. I just ran it again and it says "No balance found" so i assume that means it's done? So i think i'll wait for the parity check to complete and then remove the old SSD from the cache pool. 
 

Only thing i'm confused about: Why does my cache say "size: 1.5TB" with 3x1TB drives and the "used" and "free" numbers add up to 1.09TB? Is this normal?

[JorgeB]

3 minutes ago, YourNightmar3 said:

I just ran it again and it says "No balance found" so i assume that means it's done?

Should be but would prefer new diags.

 

4 minutes ago, YourNightmar3 said:

Is this normal?

It's a known btrfs issue with and odd number of devices in a pool.

[YourNightmar3]

Quote

It's a known btrfs issue with and odd number of devices in a pool.

Ah okay thank you

 

Quote

Should be but would prefer new diags.

 

I have attached new diagnostics.

tower-diagnostics-20221213-1348.zip

[JorgeB]

Looks good 👍