I have used two 250GB SSD's as cache drives for quite a while. Both were mounted in a RAID0 enclosure (Raidon MR2020).

 

When I was rebuilding my server I planed to take advantage of the Cache Pool feature of unRAID 6. However for the sake of simplicity I just started with on of the SSD's. As everything is going really well now (Plugins, Docker, VM's in all flavors) I would like to finally add the 2nd SSD for security reason (mirroring).

 

The drive is actually shown as BTRFS formatted and unsigned drive.

 

Can I just stop the array, attach the "new" SSD" to the Cache Pool? Anything to consider?  Thank you very much.

If I understand correctly the SSD you want to add is already BTRFS formatted, if so I would preclear it first, then just add it to the cache pool and let Unraid build mirror, you can watch progress on the cache webpage, balance status.

Thanks for your quick reply Johnnie. I started the preclear process but I don't know if is a good idea to preclear an SSD? It is really running slow at 5 MB/s. 11GB were written in 37min.

There's no problem precleaing SSDs, I do it all the time, no issues and speed is fast, usually the write speed of that particular SSD, but if you cancel now it should be enough, as long as the SSD is no longer indentified as BTFRS formatted it can be safely added to the pool.

EdgarWallace, you don't want to preclear a SSD.  You might ruin it.  I would stop the preclear ASAP.

EdgarWallace, you don't want to preclear a SSD.  You might ruin it.  I would stop the preclear ASAP.

 

I’m sorry but this it false, preclearing an SSD won’t cause any more wear than any other write operation, modern SSDs can writhe many TBs so running one preclear cycle will at most make a negligible difference on lifespam, I have a test server with SSDs only, some of them have been through more than 10 preclears and they are not “ruined”.

EdgarWallace, you don't want to preclear a SSD.  You might ruin it.  I would stop the preclear ASAP.

 

I’m sorry but this it false, preclearing an SSD won’t cause any more wear than any other write operation, modern SSDs can writhe many TBs so running one preclear cycle will at most make a negligible difference on lifespam, I have a test server with SSDs only, some of them have been through more than 10 preclears and they are not “ruined”.

 

I think the better question is, does preclearing an SSD actually help like preclearing an HDD would?

I think the better question is, does preclearing an SSD actually help like preclearing an HDD would?

 

I believe it was a fast / simple option for this case, to clear the BTRFS partition, don’t believe is needed for normal SSD use or testing like HDDs.

 

I use it many times in my test server to return SSDs to clear state so I can run parity checks with any number of disks and not needing to recalculate parity.

 

Thanks guy's for your help, particularly to johnnie.black; luckily nothing was damaged  I precleared the drive and was just adding it to the cache pool:

Label: none  uuid: a86479eb-d9bd-48a1-83db-a68fde6532e1
Total devices 2 FS bytes used 109.06GiB
devid    1 size 232.89GiB used 182.03GiB path /dev/sdb1
devid    2 size 232.89GiB used 182.03GiB path /dev/sdc1

btrfs-progs v4.1.2

Data, RAID1: total=180.00GiB, used=108.88GiB
System, RAID1: total=32.00MiB, used=48.00KiB
Metadata, RAID1: total=2.00GiB, used=182.36MiB
GlobalReserve, single: total=64.00MiB, used=0.00B

 

I was running a BALANCE and the log is showing a significant amount of these messages:

Dec 17 23:08:25 Tower kernel: BTRFS info (device sdb1): relocating block group 454759022592 flags 17
Dec 17 23:08:28 Tower kernel: BTRFS info (device sdb1): found 1000 extents
Dec 17 23:08:37 Tower kernel: BTRFS info (device sdb1): found 1000 extents
Dec 17 23:08:37 Tower kernel: BTRFS info (device sdb1): relocating block group 453685280768 flags 17
Dec 17 23:08:40 Tower kernel: BTRFS info (device sdb1): found 209 extents
Dec 17 23:08:50 Tower kernel: BTRFS info (device sdb1): found 209 extents
Dec 17 23:08:51 Tower kernel: BTRFS info (device sdb1): relocating block group 452611538944 flags 17
Dec 17 23:08:53 Tower kernel: BTRFS info (device sdb1): found 182 extents
Dec 17 23:09:05 Tower kernel: BTRFS info (device sdb1): found 182 extents

 

WebGUI is showing: No balance found on '/mnt/cache' (I left the options unchanged: "-dconvert=raid1 -mconvert=raid1").

 

Is that a correct status?

Pool status looks normal, believe those logs entries are normal during a balance operation.

 

I never needed to do a manual balance, adding a disk to pool and starting array triggers the balance, it's normal to see "No balance found on '/mnt/cache'" when not doing a balance operation.

 

You can confirm that everything is ok by running a scrub, should return 0 errors.

The manual balance was just to see what happens, it was certainly not needed. Thanks for your confirmation that all is looking fine. Scrub was indeed reporting 0 errors. Thanks again johnnie.black

Recommend doing regular scrubs to check pool as I found that the main page does not usually report write errors or red balls pool disks, a pool disk can drop offline and it will only be reported on main page after reboot or by looking at the log.

Great new feature for unraid.  Does the cache pool have to be identical drives?  Or will it work like the rest of the system, JBOD style?  I have several SSD's lying around, would be nice to put them together...