You can have dual disk failures, if configured correctly.
Data is stored on the individual disks in a standalone file system XFS, so in worst case scenario with multi-disk failure, you lose what's on those disks, instead of the entire array.
A wholllle lot less stressful when it comes to rebuilding a disk.
And you get other functionality like docker and VM support, if you want to go there.
I don't work for them. But I use their software, and I couldn't imagine using conventional RAID ever again.
100% agree with you on unRAID. I have an unRAID server with a bunch of random sized HDDs. 2 drive parity set. Actually in the process of taking a 16TB out of the parity pool and adding it to the storage pool by replacing it with a 20TB HDD. Love not having to think too hard about adding more storage to the array. Just had to think a little bit when I ran out of physical space in my tower and had to upgrade to a bigger tower that could hold 16x 3.5" HDDs along with some nice placeholders for 2.5" SSDs behind the motherboard. Anyway, yeah unRAID makes all of this fun and easy.. well aside from the OS being on a thumb drive. I've had to replace the thumb drive from failures a few times now, and no I didn't buy cheap ones.
I used Unraid a decade ago and it was kind of crap - it had little customization and not a lot of features like 'snapshots' or selective rebuild. Unraid has come so far since and I have three licenses between my home and my office.
RAID 10 does not make any sense in that configuration. RAID 6 will provide the same capacity and can withstand any 2 failures - not just specific pairs. Rebuilds are a rare event. Not losing data is paramount.
The rebuild of an array of these 22 TB drives would require about thirty-plus hours, if the system can sustain 200 MB / per-second / per-disk, up to multiples of that, if the disks are slower or the job is disturbed by other workloads, and of course, all is depending on the algorithm too. On a busy array a full —online— rebuild can easily take weeks.
I'm sure everyone knows this, but it's worth repeating.
The scary part about rebuilding an array or drive (including in unraid) is the stress it puts on every other drive in the system.
Arrays are regularly built from the same manufacturing batch of drives, which is a huge problem when one fails. It means the others are now likely in the "ready to fail" time frame as well.
If your array can't tolerate another failure, it's a very long nail-biting rebuild.
unRAID makes that better, as you never lose everything in your array in the event of a catastrophic multi-disk failure. And, hopefully, since you can mix and match drive sizes (with some limitations), you're less likely to have all drives from the same manufacturing batch.
Regardless, in all cases, an array of any sort (raid or unraid) is not a replacement for a backup.
I use two unraids for the reason. One online, one off, and once a month I manually sync the offline one to the online one. Manually as it would allow me to catch things like user error or ransomeware.
Third consecutive article referencing 'boo access'....; impressive proofreading/spellchecking software, and, perusal of comments. (I've pointed it out 3 times now..; I'm sure the next drive/SSD review....it'll make another appearance!)
"Booting Windows 10 The read-write bandwidth recorded for each drive in the ****boo access*** trace is presented below."
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
15 Comments
Back to Article
charlesg - Tuesday, January 2, 2024 - link
UnRaid is your friend.You can have dual disk failures, if configured correctly.
Data is stored on the individual disks in a standalone file system XFS, so in worst case scenario with multi-disk failure, you lose what's on those disks, instead of the entire array.
A wholllle lot less stressful when it comes to rebuilding a disk.
And you get other functionality like docker and VM support, if you want to go there.
I don't work for them. But I use their software, and I couldn't imagine using conventional RAID ever again.
BioHazardous - Tuesday, January 2, 2024 - link
100% agree with you on unRAID. I have an unRAID server with a bunch of random sized HDDs. 2 drive parity set. Actually in the process of taking a 16TB out of the parity pool and adding it to the storage pool by replacing it with a 20TB HDD. Love not having to think too hard about adding more storage to the array. Just had to think a little bit when I ran out of physical space in my tower and had to upgrade to a bigger tower that could hold 16x 3.5" HDDs along with some nice placeholders for 2.5" SSDs behind the motherboard. Anyway, yeah unRAID makes all of this fun and easy.. well aside from the OS being on a thumb drive. I've had to replace the thumb drive from failures a few times now, and no I didn't buy cheap ones.Samus - Wednesday, January 3, 2024 - link
I used Unraid a decade ago and it was kind of crap - it had little customization and not a lot of features like 'snapshots' or selective rebuild. Unraid has come so far since and I have three licenses between my home and my office.Threska - Wednesday, January 3, 2024 - link
Xigma uses ZFS as well.https://xigmanas.com/xnaswp/
charlesg - Wednesday, January 3, 2024 - link
unRAID support ZFS now as well.ballsystemlord - Tuesday, January 2, 2024 - link
You could also try RAID 6. It's what servers use from what I understand. And it's quite robust.awolfe63 - Tuesday, January 2, 2024 - link
RAID 10 does not make any sense in that configuration. RAID 6 will provide the same capacity and can withstand any 2 failures - not just specific pairs. Rebuilds are a rare event. Not losing data is paramount.meacupla - Tuesday, January 2, 2024 - link
How long would it take to rebuild a RAID 6 array made from 22TB drives?Tunnah - Wednesday, January 3, 2024 - link
Some mice built a computer that worked faster..DrReD - Wednesday, January 3, 2024 - link
The rebuild of an array of these 22 TB drives would require about thirty-plus hours, if the system can sustain 200 MB / per-second / per-disk, up to multiples of that, if the disks are slower or the job is disturbed by other workloads, and of course, all is depending on the algorithm too.On a busy array a full —online— rebuild can easily take weeks.
charlesg - Wednesday, January 3, 2024 - link
I'm sure everyone knows this, but it's worth repeating.The scary part about rebuilding an array or drive (including in unraid) is the stress it puts on every other drive in the system.
Arrays are regularly built from the same manufacturing batch of drives, which is a huge problem when one fails. It means the others are now likely in the "ready to fail" time frame as well.
If your array can't tolerate another failure, it's a very long nail-biting rebuild.
unRAID makes that better, as you never lose everything in your array in the event of a catastrophic multi-disk failure. And, hopefully, since you can mix and match drive sizes (with some limitations), you're less likely to have all drives from the same manufacturing batch.
Regardless, in all cases, an array of any sort (raid or unraid) is not a replacement for a backup.
I use two unraids for the reason. One online, one off, and once a month I manually sync the offline one to the online one. Manually as it would allow me to catch things like user error or ransomeware.
ballsystemlord - Wednesday, January 3, 2024 - link
Yes, cold storage is important.MDD1963 - Wednesday, January 3, 2024 - link
Third consecutive article referencing 'boo access'....; impressive proofreading/spellchecking software, and, perusal of comments. (I've pointed it out 3 times now..; I'm sure the next drive/SSD review....it'll make another appearance!)"Booting Windows 10
The read-write bandwidth recorded for each drive in the ****boo access*** trace is presented below."
ganeshts - Wednesday, January 3, 2024 - link
The official PCMark 10 Storage Bench documentation refers to the access trace as 'boo':https://support.benchmarks.ul.com/support/solution...
ballsystemlord - Wednesday, January 3, 2024 - link
So, it's actually PCMark 10's spell checker problem.