v6.9.0 - Cache Pools with different SSD disk sizes

[FlorinB]

Background: Initially I had a Cache Pool with 2 x 1TB SSD disks. With Unraid v.6.9.0 it is possible to bulid multiple pools. 

 

I have added to my Unraid another 2 SSDs - one of 1TB and the second of 2TB.

 

By using the btrfs calculator from https://carfax.org.uk/btrfs-usage/ results that if I add 1TB+2TB in a raid, even with RAID0 i will lose 1 TB.

 

 

1. Should I add the 1TB disk to the Cache pool as the 3rd disk and build the additional pool with 2TB disk only?

 

2. If I build the second pool with only one disk how can I have at least Metadata 2nd copy on the same disk? Because at this moment everything is set on sigle.

Data, single: total=1.00GiB, used=0.00B
System, single: total=32.00MiB, used=16.00KiB
Metadata, single: total=1.00GiB, used=112.00KiB
GlobalReserve, single: total=3.25MiB, used=0.00B

 

3. What other options do I have?

 

Best regards,

Florin

 

 

 

 

[JorgeB]

5 minutes ago, FlorinB said:

2. If I build the second pool with only one disk how can I have at least Metadata 2nd copy on the same disk? Because at this moment everything is set on sigle.

By default rotational devices are formatted with dup metadata by btrfs, non rotational with single, there's some discussion if it's really worth having dup metadata on non rotational devices, but you can convert manually, I use dup metadata even for SSDs.

 

btrfs balance start -mconvert=dup /mnt/pool

 

[JorgeB]

8 minutes ago, FlorinB said:

3. What other options do I have?

If you don't need redundancy you could use both devices with the single profile (with or without dup metadata) to fully use the 3TB.

[FlorinB]

5 minutes ago, johnnie.black said:

By default rotational devices are formatted with dup metadata by btrfs, non rotational with single, there's some discussion if it's really worth having dup metadata on non rotational devices, but you can convert manually, I use dup metadata even for SSDs.

 

btrfs balance start -mconvert=dup /mnt/pool

 

Now I have:

 

root@Node804:~# btrfs fi df /mnt/pool_2tb/
Data, single: total=1.00GiB, used=0.00B
System, DUP: total=32.00MiB, used=16.00KiB
Metadata, DUP: total=1.00GiB, used=112.00KiB
GlobalReserve, single: total=3.25MiB, used=0.00B

 

and into the GUI is displaying:

Is this what I should expect? Do i still have the 2 TB available for data?

 

 

[JorgeB]

Just now, FlorinB said:

Is this what I should expect?

Yep, it only duplicates the metadata, not the data.

[FlorinB]

I have read the below thread. If there are minimum 2 devices it is recommended to use always  -mconvert=raid1 for metadata redundancy?

btrfs balance start -dconvert=... -mconvert=raid1 /mnt/cache

 

 

[JorgeB]

1 minute ago, FlorinB said:

If there are minimum 2 devices it is recommended to use always  -mconvert=raid1 for metadata redundancy?

With 2 devices yes, use raid1, dup is mostly for single device use.

[FlorinB]

@johnnie.black and the last questions:

 

1. Could you please suggest when it will be better to have redundancy on the Pools and for which kind of data.

2. What would be some use cases for having multipe Pools. I understand that adding SSDs into the array at this moment is not possible.

3. In a 2 or more btrfs disks Pool, assuming that it is at least raid1, if one disk fails what are the chances to get the data from there? Is it better to copy all data in another place and rebuild the Pool from scratch or just add the replacement disk?

 

Thank you very much for your help and quick replies Johhnie.

[JorgeB]

12 minutes ago, FlorinB said:

Could you please suggest when it will be better to have redundancy on the Pools and for which kind of data.

That's really up to you, also note that redundancy not the same as backup, I have multiple pools and they are almost all redundant, still do regular backups of anything important (usually to another pool on a different server).

 

14 minutes ago, FlorinB said:

What would be some use cases for having multipe Pools. I understand that adding SSDs into the array at this moment is not possible.

Again it really depends on your use case, I use them mostly to keep things separate and also for when a pool is being heavily used it won't affect performance of the others.

 

21 minutes ago, FlorinB said:

3. In a 2 or more btrfs disks Pool, assuming that it is at least raid1, if one disk fails what are the chances to get the data from there?

Good chance if it's really a failing disk, some intermittent errors where disks drop offline and came back online can cause issues with btrfs pools, if it fails you can just replace the disk, if there are more serious issues restore from backup.

 

 

[FlorinB]

I am considering to use a 3 x 1TB btrfs pool in raid5. Readed about the "write hole", but there are voices which are saying that with a UPS it will be safe enough

 

https://blog.wille.io/index.php/2018/12/28/btrfs-with-raid5-is-safe-now/

 

Any oppinions on this matter?

Is someone else using btrfs with raid5 and for how much time?

Any issues occurred so far?

Edited July 14, 2020 by FlorinB

[JorgeB]

It's considered safe enough for most if running raid1 metadata, Unraid default, should not use raid5/6 for metadata, I'm running several raid5 pools for more than a year without issues, still should have backups of anything important.