And yes, I already know the RAID 0 disadvantages where when one of the
disk fails, all data is lost. But then, this is RAID 5 on top.
assuming the 500GB fails, I could just slap another 500GB disk there,
and then rebuild everything from the RAID 5. Right? Right?
Be very specific with your recovery procedure. Replace the physical disk in the RAID 0 array. However RAID 0 cannot be rebuilt, there will be corruption striped in it. Remove and re-add the replaced RAID 0 member of the RAID 5 array, not the physical drive.
Introduces a little operational complexity. Beginner how to repair an array articles are focused on physical drives, not nested arrays. Also, not all array software can do RAID 05. Linux md software raid should allow arbitrary nested, possibly some enterprise arrays are capable as well.
Now I want to make RAID 5 out of (RAID 0 2x 1TB), (RAID 0 4x500GB),
and 2TB disk.
Theoretically you can, but in practice don't do that.
Any member of RAID 0 failing is a problem. Odds of success decrease as more disks are added. Say a certain drive model can expect one unrecoverable error in 1e14 bytes read. RAID 0 with 4x of these means all need to independently achieve this with 99.999999999999% chance of success. Multiply the probabilities together, and pretty soon you get 13 nines of durability rather than 14.
Which is a part of a redundant array, yes, but this introduces unnecessary risk of degrading that array. Your layout is less reliable than if it was a traditional 7 disk RAID 6. Further, double parity is superior, maintains fault tolerance even during rebuild.
Admittedly the 05 of varying sizes is more space efficient. Although I would argue that the operational simplicity of all drives the same size is worth the cost of sourcing them.
