Exactly, at that point it has all the disadvantages of RAIDZ2, with all the disadvantages of DRAID, with none of the benefits.
1 Like
What are the disadvantages? I saw IOPS being mentioned but haven’t researched exactly how big of a hit.
Interesting. I saw another article that implied anything around 20 drives.
Obviously I’m no where near 100 drives… so I guess raidz2 12-wide / 2-vdev would be better for me… I can’t justify raidz3 beyond a single vdev.
The main disadvantage is that all allocations in a DRAID vdev are multiples of the stripe width times the ashift. That means that, if you need to allocate a byte, you need to allocate 12 chunks (10 “data disks” plus two “parity disks” in your example). With RAIDZ2, you’d allocate three (1 data + 2 parity) in the same situation.
4 Likes
Interesting… Well, I ended up rebuilding as a RAIDz2 w/ 12-wide x 2 vdevs.
I still ended up with 365TiB of usable space… Doesn’t that seem like a lot of overhead? I would have expected ~400 TiB usable space.
It’s all just an estimate, anyway, but I do agree that the space reporting seems more conservative than it used to be
1 Like
Aha, this is the first time I’ve read that it’s just an estimate. Is that because of the variable block sizes and compression? Does it ‘grow’ over time as the drive fills?
Variable block sizes (so different metadata to data ratios, and different parity to data ratios for RAIDZ) mostly. Conceptually, compression will also have an impact, but that’s not accounted for in the free space.
Maybe? Again, there was recent work around this reporting and I’m not sure what the latest behavior is.
In practical terms, the “available” value will decrease more slowly than the logical volume of data (e.g. as reported by du) grows, if your data compresses.
1 Like
If such a thing happened to me I would freak out and scramble for my backup.
What TN tells you re: used space is NOT an estimation. What we and other websites may tell you IS. TN number is final.
There is a TN official calculator somewhere in the documentation.
It is absolutely an estimate, because nobody knows a priori what the data looks like and how it will end up stored on RAID.
As far as I understand the number of used space TN tells you is absolute.
Ah, yes, used space is unambiguous and a real measurement. I’m talking about the “available” figure.
1 Like
At what point does the available figure ‘update’? I’ve used 310TiB with 57TiB currently available. Which is right on track for the 366T of available space my pool has. I’m still wondering where the ~40TiB of space went…sigh.
Between the 4 parity drives (2x in each vdev) and the “overhead”, I’m out about 114TiB of space(!). Is this expected? Am I the only one who thinks something seems wrong here?
Thanks, as long as I’m not overlooking something!