Hi guys,

After going through encrypting my data drives and moving data back and forth, I'm ready for the next step- cache drive encryption.

I have read some discussion on it, including the fact that TRIM is not supported at the dm-layer, so overprovisioning is the only option at the moment.

 

I currently have the following:

1. 512Gb cache SSD drive with dockers only formatted btrfs

2. Empty 512Gb SSD drive, newly purchased, not added to the config

 

Desired outcome: 2 mirrored cache drive formatted btrfs-enc

 

Could somebody be so kind to list the steps of what needs to be done?

 

Thank you

-Backup cache

-Add the other device to pool and format with btrfs encrypted

-Restore data to cache

 

You can use this to help with the backup/restore:

https://forums.unraid.net/topic/46802-faq-for-unraid-v6/?do=findComment&comment=511923

 

And what about TRIM and overprovisioning?

And what about TRIM and overprovisioning?

Can't help with that, I don't use encryption so never had to worry about it, but google Unraid HPA, it should work if you create one, but before doing it use blkdiscard or secure erase on the SSDs.

 

 

I can't believe nobody's done it before?

So open topics:

1. What's the consensus on overprovisioning of encrypted SSD cache drives?

2. Should I pre-clear a SSD?

2 hours ago, Kir said:

2. Should I pre-clear a SSD?

No, use:

 blkdiscard /dev/sdX

 

blkdiscard: /dev/sdi: BLKDISCARD ioctl failed: Remote I/O error

 

In log:

Dec 18 16:31:12 Storage kernel: sd 1:0:8:0: [sdi] tag#3 UNKNOWN(0x2003) Result: hostbyte=0x00 driverbyte=0x08
Dec 18 16:31:12 Storage kernel: sd 1:0:8:0: [sdi] tag#3 Sense Key : 0x5 [current]
Dec 18 16:31:12 Storage kernel: sd 1:0:8:0: [sdi] tag#3 ASC=0x21 ASCQ=0x0
Dec 18 16:31:12 Storage kernel: sd 1:0:8:0: [sdi] tag#3 CDB: opcode=0x42 42 00 00 00 00 00 00 00 18 00
Dec 18 16:31:12 Storage kernel: print_req_error: critical target error, dev sdi, sector 998244232

That means the SSD is on a controller that doesn't support trim, use onboard SATA instead (set to AHCI)

Following this thread with interest. 

I just moved all my drives over to encrypted fs, mostly to protect the data if hw stolen. 

I have 2 SSDs as cache and fs is encrypted btrfs in raid 1. Trim "should" work even on encrypted drives, right? I understand it won't on unraid (yet?) but what problems can this cause me? Over provisioning is an acceptable workaround if it helps with anything. But how do we do that?

 

Would be nice to be able to set size when formatting the drives in the gui. Just leave some unformatted is what over provisioning does? 

 

 

Edited December 19, 20187 yr by Niklas

1 hour ago, Niklas said:

have 2 SSDs as cache and fs is encrypted btrfs in raid 1. Trim "should" work even on encrypted drives, right?

No, you should get an unsupported operation error.

 

1 hour ago, Niklas said:

Over provisioning is an acceptable workaround if it helps with anything. But how do we do that?

 

On 12/9/2018 at 12:25 AM, johnnie.black said:

google Unraid HPA, it should work if you create one, but before doing it use blkdiscard or secure erase on the SSDs.

 

15 hours ago, johnnie.black said:

That means the SSD is on a controller that doesn't support trim, use onboard SATA instead (set to AHCI)

It's connected to the same controller as the other SSD, and I'm seeing this in the syslog:

root: /var/lib/docker: 6.8 GiB (7294545920 bytes) trimmed

6 minutes ago, Kir said:

root: /var/lib/docker: 6.8 GiB (7294545920 bytes) trimmed

That's the docker image getting trimmed, not the SSD filesystem.

Just now, johnnie.black said:

That's the docker image getting trimmed, not the SSD filesystem.

Damn, and I always thought it was working 😕

Since I can't move it to another controller (Supermicro cage), is there another command to issue?

You can preclear it, though it will cause some wear on the SSD, but doing it one time is not a big deal.

I managed to do secure erase instead following this guide