Upgrading Cache Pool v6.7.0

[CowboyRedBeard]

I'm on 6.7.0 and as a part of my migration upward in releases and expanding the array... I've also gotten larger cache drives.

 

Presently I have 2x 500G SSD

 

I've gotten 2x 1TB SSD to expand.

 

If I understand correctly:

 

1 - Shut down the server.

2 - Replace 2nd ("Cache 2") drive with a new larger one.

3 - Restart

4 - After it's had time to join that drive into the pool, shut down again

5 - Replace 1st cache drive with a new larger one

6 - Restart

 

Is that right?

[JonathanM]

As long as your current pool is healthy and properly RAID1, I think that is correct. However... I think the chances are pretty high your current pool may need some work before you proceed.

 

Attach your diagnostics zip file to your next post in this thread, and wait for @johnnie.black to weigh in on your next steps.

 

However you proceed, I recommend being sure your backups for significant files in the pool are current before moving on.

[JorgeB]

Just make sure pool is actuality redundant before starting, i.e., it won't be if it was created on v6.7.x due to this bug.

[CowboyRedBeard]

How do I verify that it's redundant? It was created before 6.7....

 

On Main it says 

Device is part of cache pool

For the second disk

[JorgeB]

Post diags or output of:

 

btrfs fi usage -T /mnt/cache

 

[CowboyRedBeard]

root@Tower:~# btrfs fi usage -T /mnt/cache
Overall:
    Device size:		 931.52GiB
    Device allocated:		 216.06GiB
    Device unallocated:		 715.46GiB
    Device missing:		     0.00B
    Used:			 109.29GiB
    Free (estimated):		 409.52GiB	(min: 409.52GiB)
    Data ratio:			      2.00
    Metadata ratio:		      2.00
    Global reserve:		 117.86MiB	(used: 0.00B)

             Data      Metadata  System              
Id Path      RAID1     RAID1     RAID1    Unallocated
-- --------- --------- --------- -------- -----------
 2 /dev/sdb1 106.00GiB   2.00GiB 32.00MiB   357.73GiB
 1 /dev/sdc1 106.00GiB   2.00GiB 32.00MiB   357.73GiB
-- --------- --------- --------- -------- -----------
   Total     106.00GiB   2.00GiB 32.00MiB   715.46GiB
   Used       54.21GiB 445.72MiB 16.00KiB            

 

[JorgeB]

It's fine, you can proceed.

[CowboyRedBeard]

So I've completed replacing the 2nd cache drive with the new one, but it's saying the pool is now 1.42T (the old drive is 500G and the new one is 1T). I had expected it to say 500G for the pool size.

 

So am I now not redundant? Just afraid if I replace the 1st drive (old 500G) that it'll blow up?

 

root@Tower:~# btrfs fi usage -T /mnt/cache
Overall:
    Device size:		   1.36TiB
    Device allocated:		  75.06GiB
    Device unallocated:		   1.29TiB
    Device missing:		     0.00B
    Used:			  73.28GiB
    Free (estimated):		   1.29TiB	(min: 661.70GiB)
    Data ratio:			      1.00
    Metadata ratio:		      2.00
    Global reserve:		 117.66MiB	(used: 0.00B)

             Data     Metadata  System              
Id Path      single   RAID1     RAID1    Unallocated
-- --------- -------- --------- -------- -----------
 3 /dev/sdb1 73.00GiB   1.00GiB 32.00MiB   857.48GiB
 1 /dev/sdc1        -   1.00GiB 32.00MiB   464.73GiB
-- --------- -------- --------- -------- -----------
   Total     73.00GiB   1.00GiB 32.00MiB     1.29TiB
   Used      72.41GiB 448.80MiB 16.00KiB            

 

[JorgeB]

15 minutes ago, CowboyRedBeard said:

So am I now not redundant?

Currently not, data is single profile, you can convert back to raid1, but it should have stayed raid1 during and after the upgrade, something is not right, if you still didn't reboot post diags.

[CowboyRedBeard]

I haven't rebooted since, but when it came back online I had forgotten to not auto-start the array. So it was giving me "btrfs operation " next to where the stop array button was. When I checked the logs it was doing this:

 

Mar 23 12:40:28 Tower kernel: BTRFS info (device sdc1): found 4 extents
Mar 23 12:40:42 Tower kernel: BTRFS info (device sdc1): found 4 extents
Mar 23 12:40:43 Tower kernel: BTRFS info (device sdc1): found 4 extents
Mar 23 12:40:43 Tower kernel: BTRFS info (device sdc1): relocating block group 5588914274304 flags data

 

Edited March 24, 2020 by CowboyRedBeard

[CowboyRedBeard]

a

 

Edited March 25, 2020 by CowboyRedBeard

[JorgeB]

12 hours ago, CowboyRedBeard said:

I haven't rebooted since, but when it came back online I had forgotten to not auto-start the array.

OK, so that's normal, when you started the array with a single cache device the pool is converted to single profile, and because of the bug with v6.7 it's not automatically convert back when a new device is added, you should also update to v6.8.3.

[CowboyRedBeard]

OK, upgraded and rebooted and on 6.8.3, looks the same:

 

root@Tower:~# btrfs fi usage -T /mnt/cache
Overall:
    Device size:                   1.36TiB
    Device allocated:            123.06GiB
    Device unallocated:            1.24TiB
    Device missing:                  0.00B
    Used:                         81.71GiB
    Free (estimated):              1.28TiB      (min: 677.30GiB)
    Data ratio:                       1.00
    Metadata ratio:                   2.00
    Global reserve:              128.64MiB      (used: 0.00B)

             Data      Metadata  System
Id Path      single    RAID1     RAID1    Unallocated
-- --------- --------- --------- -------- -----------
 1 /dev/sdc1         -   1.00GiB 32.00MiB   464.73GiB
 3 /dev/sdb1 121.00GiB   1.00GiB 32.00MiB   809.48GiB
-- --------- --------- --------- -------- -----------
   Total     121.00GiB   1.00GiB 32.00MiB     1.24TiB
   Used       80.80GiB 464.33MiB 16.00KiB

 

[JorgeB]

You still need to convert to raid1, with 6.8.3 there's a GUI option on the cache page.

[CowboyRedBeard]

All set, thanks for the help!