Dirty secret that they won't tell you: in this day and age, anything other than RAID6 (or RAIDZ2) is pretty much useless with modern drive sizes. Unless you're just striping and using some other backup scheme to protect your data.
Don't waste your time getting this and using RAID5. It's a false sense of security that you'll learn the hard way when you most need it.
Dirty secret that they won't tell you: in this day and age, anything other than RAID6 (or RAIDZ2) is pretty much useless with modern drive sizes. Unless you're just striping and using some other backup scheme to protect your data.
Don't waste your time getting this and using RAID5. It's a false sense of security that you'll learn the hard way when you most need it. - sremick
Please, do explain why RAID5 is a waste of time? I've been using it in a 4-disk MediaSonic box for a while now and each time a single disk fails I'm happy to replace it and have zero loss of data. What scenario are you commenting on that is a false sense of security? I'm asking genuinely, as I wonder if I've overlooked something. Thanks in advance!
Please, do explain why RAID5 is a waste of time? I've been using it in a 4-disk MediaSonic box for a while now and each time a single disk fails I'm happy to replace it and have zero loss of data. What scenario are you commenting on that is a false sense of security? I'm asking genuinely, as I wonder if I've overlooked something. Thanks in advance! - cyberwolf
The issue comes down to the re-silvering process after a failed drive is replaced. With modern drive capacities, the MTBF means that you're highly likely to get a read-failure during the resilvering... problem is, at that moment you now have no data redundancy, so data is lost.
It's a numbers game, and I'm glad your experiences so far have been lucky. I'm an IT professional who also moonlights on the side for home users and small businesses, and I've seen many mirrored pairs and RAID5 arrays fail during replacement for what can be chalked up to this very reason. If you search for "RAID5 is dead" you'll find no shortage of other discussions on the internet (along with detractors, but note that lucky anecdotal experiences don't make the issue any less real).
I use RAIDZ2 at home (2-drive redundancy) for this very reason. Not cheap, but if you're not doing it in a way that works then what's the point? It's like having a pretty fire extinguisher... that's empty.
^ No strategy including RAIDZ2 is as comprehensive as the direction you're trying to go. It won't protect against catastrophic events or soft errors.
Only an offline backup will protect against the latter and make that off site for best protection against the former, and once you have that offline backup, RAID5 is fine.
If we were talking about some mission critical system where there is a great financial loss from loss of productivity during down time, if the HAS goes down, you'd then have a redundant NAS too, to factor for HAS failure in addition to HDD failures.
^ No strategy including RAIDZ2 is as comprehensive as the direction you're trying to go. It won't protect against catastrophic events or soft errors.
Only an offline backup will protect against the latter and make that off site for best protection against the former, and once you have that offline backup, RAID5 is fine.
If we were talking about some mission critical system where there is a great financial loss from loss of productivity during down time, if the HAS goes down, you'd then have a redundant NAS too, to factor for HAS failure in addition to HDD failures. - dave_c
This is incorrect. RAID6/RAIDZ2 does protect against soft errors while replacing a single drive because you still have redundancy during the resilvering. Even RAID5 protects against soft read-errors during normal operations. The only situation where you'd be vulnerable to soft errors with RAID6/RAIDZ2 would be if you're replacing two failed drives simultaneously.
I am not disagreeing that RAID != backup, but people use RAID so that they don't have to resort to the backup (which may be slow, inconvenient, and/or offsite) for every little thing. If you have to resort to the backup often when replacing a failed drive in a RAID5 array, it's not serving its intended purpose very well.
Not clear where you're getting your incorrect understanding of RAID and soft errors from.
^ You're trying to ignore the other sources of soft errors including hardware, OS, and malware caused. The data is already dirty or gone so successfully replacing the drive won't matter.
^ You're trying to ignore the other sources of soft errors including hardware, OS, and malware caused. The data is already dirty or gone so successfully replacing the drive won't matter. - dave_c
If you have all those other things going on, you have much bigger problems in your design that you need to address.
You start with reliable drives, but because all drives are expected to have occasionally soft-errors you use a redundant file system with error checking, such as ZFS.
You protect against RAM-induced soft errors by using ECC RAM.
You protect against OS soft errors by using a reliable OS. There should NOT be an "acceptable" or "norm" of soft errors here, therefore if you are finding yourself facing them the solution isn't to harden against them, it's time to use a different OS.
Altogether this cons***utes a quality NAS.
(edit: BB, your censorship is boneheaded)
"Malware" is in the category of "user induced".... no different than "oh *** I deleted a file I didn't mean to". A filesystem like ZFS can be configured to use snapshots which can help, but you and I are in agreement that RAID != backup and so it's important to have a backup regardless. What we disagree is what is the norm that one should have to utilize the backup. The purpose of RAID is continuity to smoothly survive a drive failure. If you have to resort to the backup frequently to recover from a single drive failure, your RAID is not doing its job and is pointless. You should only have to resort to your backup for A) catastrophic total loss of the NAS, and B) client-induced errors (malware, user, etc) beyond the control of the NAS.
Comments & Reviews (8)
Don't waste your time getting this and using RAID5. It's a false sense of security that you'll learn the hard way when you most need it.
Don't waste your time getting this and using RAID5. It's a false sense of security that you'll learn the hard way when you most need it. - sremick
It's a numbers game, and I'm glad your experiences so far have been lucky. I'm an IT professional who also moonlights on the side for home users and small businesses, and I've seen many mirrored pairs and RAID5 arrays fail during replacement for what can be chalked up to this very reason. If you search for "RAID5 is dead" you'll find no shortage of other discussions on the internet (along with detractors, but note that lucky anecdotal experiences don't make the issue any less real).
I use RAIDZ2 at home (2-drive redundancy) for this very reason. Not cheap, but if you're not doing it in a way that works then what's the point? It's like having a pretty fire extinguisher... that's empty.
Only an offline backup will protect against the latter and make that off site for best protection against the former, and once you have that offline backup, RAID5 is fine.
If we were talking about some mission critical system where there is a great financial loss from loss of productivity during down time, if the HAS goes down, you'd then have a redundant NAS too, to factor for HAS failure in addition to HDD failures.
Only an offline backup will protect against the latter and make that off site for best protection against the former, and once you have that offline backup, RAID5 is fine.
If we were talking about some mission critical system where there is a great financial loss from loss of productivity during down time, if the HAS goes down, you'd then have a redundant NAS too, to factor for HAS failure in addition to HDD failures. - dave_c
I am not disagreeing that RAID != backup, but people use RAID so that they don't have to resort to the backup (which may be slow, inconvenient, and/or offsite) for every little thing. If you have to resort to the backup often when replacing a failed drive in a RAID5 array, it's not serving its intended purpose very well.
Not clear where you're getting your incorrect understanding of RAID and soft errors from.
You start with reliable drives, but because all drives are expected to have occasionally soft-errors you use a redundant file system with error checking, such as ZFS.
You protect against RAM-induced soft errors by using ECC RAM.
You protect against OS soft errors by using a reliable OS. There should NOT be an "acceptable" or "norm" of soft errors here, therefore if you are finding yourself facing them the solution isn't to harden against them, it's time to use a different OS.
Altogether this cons***utes a quality NAS.
(edit: BB, your censorship is boneheaded)
"Malware" is in the category of "user induced".... no different than "oh *** I deleted a file I didn't mean to". A filesystem like ZFS can be configured to use snapshots which can help, but you and I are in agreement that RAID != backup and so it's important to have a backup regardless. What we disagree is what is the norm that one should have to utilize the backup. The purpose of RAID is continuity to smoothly survive a drive failure. If you have to resort to the backup frequently to recover from a single drive failure, your RAID is not doing its job and is pointless. You should only have to resort to your backup for A) catastrophic total loss of the NAS, and B) client-induced errors (malware, user, etc) beyond the control of the NAS.
Thank you!