Hi folks,

 

I have this question regarding my cache pool, I use that world maybe in a looser way then you do:

 

I don't use it exactly that way because yes, I store all my application data for my containers and such but instead of having the other files such as new downloads and docs move automatically from the cache to the array I prefer to set them to cache only, then move them manually as I see fit.

 

Now that being aside: 

 

That drive is a 1TB NVME, I have 2 other 525GB regular SSDs that I would like to add to the cache pool and use the NVME as parity for them since it has to be the biggest drive.

 

My main question is then:

 

What would be the best way to move the data on the 1TB NVME to a pool made with the 2 525GB SSDs in order to use the NVME as parity without disrupting anything? Or at least as little as possible?

 

Thanks again for letting me know if possible.

If you are on 7.0.x you can use the mover, but make sure you disable the Docker service first.

I think that after some reading, there was a misunderstanding because I may have used the wrong terms:

 

From what I understand, there is no such thing as parity calculation outside of the main array?

 

If that is the case no problem but how can I have some sort of raid1/real-time mirrorring between the nvme and the 2 500gb ssds pooled together?

 

Even better if I can achieve that by keeping the NVME as main drive so I won't need to move the appdata folders.

 

Thanks again for any help, I just want to be sure of my decisions before proceeding.

 

6 hours ago, MonadProxy said:

there is no such thing as parity calculation outside of the main array?

There isn't, but that wasn't what you were asking.

 

6 hours ago, MonadProxy said:

If that is the case no problem but how can I have some sort of raid1/real-time mirrorring between the nvme and the 2 500gb ssds pooled together?

You can't, a single device pool is always unprotected, a two device pool defaults to raid1 (mirror)

7 hours ago, MonadProxy said:

f that is the case no problem but how can I have some sort of raid1/real-time mirrorring between the nvme and the 2 500gb ssds pooled together

If you put them all in a pool formatted using BTRFS then you will have redundancy on that pool using the BTRFS variant of RAID1.  However currently the 1TB drive is currently formatted as XFS so you would first need to copy its data elsewhere to be able to change the format.

Thanks to both of you for the explanations,

 

So if I understand correctly, I can achieve that as long as all the drives are formatted to BTRFS but having different size drives in this case, how should I go about setting this up:

 

I just move the data from the NVME, reformat it to BTRFS, move the data back, and then when I add the other 2x500gb drives to that pool I can set them to act as raid1 together even though they are of smaller size? Or would I need another equally sized drive to be able to do that?

btrfs supports a mix of drive sizes in a pool.    The btrfs raid1 profile is not traditional raid1, but it will ensure that each bit of data is held on at least 2 drives.   As long as the first drive you put into a pool is formatted as btrfs you can later add drives of any size.   Google for the btrfs calculator to see how much usable space you get from any mix of drives you want to use.