NAS Backups

[mashie]

Good thing people stopped doing hardware RAID a decade ago, with software RAID in Linux I can move all the disks to a different computer all together if required.

 

[newtekie1]

If one of the drives fails and you have to rebuild, be prepared for an extremely long wait and the possibility that the rebuild fails thus losing all of your data.
I recently built a NAS and was initially looking at some form of RAID, everything I found said that RAID is a bad idea for larger drives. especially if you don't have backups!

I've rebuilt arrays with 6TB disks before, it really doesn't take that long, and a rebuild failing doesn't result in all the data on the array being lost. The array just goes back to a degraded state.

I always see people making statement like that that have never actually done it.

There are better methods of having reliable parity, other than raid

At the end of the day, parity is parity. ZFS or unRAID is not any better than a dedicated hardware RAID controller. A drive failure still results in a rebuild pretty much the same as with a RAID5/6.

Filesystems like ZFS or unRAID work better for larger disks. They will, generally, use less disks and allow you to keep your backups on the spare disks you have left over.

This statement literally makes no sense. ZFS's parity is either 1 disk or 2 disk, just like RAID5/6. It uses the same number of drives and gives the same amount of storage space.

Good thing people stopped doing hardware RAID a decade ago, with software RAID in Linux I can move all the disks to a different computer all together if required.

I can do that with my hardware RAID. Hell, I can even move it from a Windows computer to a Linux computer if I wanted...

 

[silkstone]

Good thing people stopped doing hardware RAID a decade ago, with software RAID in Linux I can move all the disks to a different computer all together if required.

And how often do you scrub that array?

I've rebuilt arrays with 6TB disks before, it really doesn't take that long, and a rebuild failing doesn't result in all the data on the array being lost. The array just goes back to a degraded state.

I always see people making statement like that that have never actually done it.

At the end of the day, parity is parity. ZFS or unRAID is not any better than a dedicated hardware RAID controller. A drive failure still results in a rebuild pretty much the same as with a RAID5/6.

This statement literally makes no sense. ZFS's parity is either 1 disk or 2 disk, just like RAID5/6. It uses the same number of drives and gives the same amount of storage space.

I can do that with my hardware RAID. Hell, I can even move it from a Windows computer to a Linux computer if I wanted...

I am by no means an expert, but literally everything I have read says to stay the hell away from raid when working with large disks and that RAID is not a backup solution. I guess it not being backup isn't important for data that isn't critical, but if it is, by having your disks in RAID rather than as a backup, you might be lulling yourself into a false sense of security regarding the redundancy of that data.

Here's some info on Raid 6 rebuild times: https://serverfault.com/questions/967930/raid-5-6-rebuild-time-calculation

Whilst it's pretty reliable, the rebuild times, when looking at arrays over 50Tb, are pretty insane. If you don't mind putting your array out of commission while it rebuilds, I guess it's a fine solution. I just read about lots of people moving away from all forms of RAID.

 

[newtekie1]

I am by no means an expert, but literally everything I have read says to stay the hell away from raid when working with large disks and that RAID is not a backup solution. I guess it not being backup isn't important for data that isn't critical, but if it is, by having your disks in RAID rather than as a backup, you might be lulling yourself into a false sense of security regarding the redundancy of that data.

If anything you read didn't tell you to keep backups for anything on the RAID, because RAID isn't a backup, then they were full of crap. But nothing I've read from any reputable source says to avoid RAID.

Here's some info on Raid 6 rebuild times: https://serverfault.com/questions/967930/raid-5-6-rebuild-time-calculation

Whilst it's pretty reliable, the rebuild times, when looking at arrays over 50Tb, are pretty insane. If you don't mind putting your array out of commission while it rebuilds, I guess it's a fine solution. I just read about lots of people moving away from all forms of RAID.

Random forum posts aren't reliable sources. Yes, I recognize the irony in saying this.

The rebuild times with ZFS or unRAID parity is just as bad as a hardware RAID. I've used them all, there really isn't a huge difference between RAID5/6 and ZFS parity, ZFS is just a software form of it. And note that ZFS is not redundant, like RAID5/6, just by default. You have to specifically enable parity(it's actually called Z-RAID) before the storage pool will become redundant.

Also, I have no idea what you mean by "if you don't mind putting your array out of commission while it rebuilds". The array is completely usable during a rebuild, or a OCE or ORLM. And most NAS devices still use RAID, it's the industry standard for redundancy.

 

[silkstone]

If anything you read didn't tell you to keep backups for anything on the RAID, because RAID isn't a backup, then they were full of crap. But nothing I've read from any reputable source says to avoid RAID.



Random forum posts aren't reliable sources. Yes, I recognize the irony in saying this.

The rebuild times with ZFS or unRAID parity is just as bad as a hardware RAID. I've used them all, there really isn't a huge difference between RAID5/6 and ZFS parity, ZFS is just a software form of it. And note that ZFS is not redundant, like RAID5/6, just by default. You have to specifically enable parity(it's actually called Z-RAID) before the storage pool will become redundant.

Also, I have no idea what you mean by "if you don't mind putting your array out of commission while it rebuilds". The array is completely usable during a rebuild, or a OCE or ORLM. And most NAS devices still use RAID, it's the industry standard for redundancy.

When the disk is rebuilding there is meant to be a huge amount of overhead slowing all access to your system or putting it out of commission for the duration of the rebuild.
Here's a (non-forum) article about the death of raid 6 in enterprises https://www.zdnet.com/article/why-raid-6-stops-working-in-2019/

 

[mashie]

And how often do you scrub that array?

Slightly more often then my previous arrays which was 0 times in 10 years.

Out of curiosity to see how long a scrub would take I did one the other day, it took 14h. If I'm bored enough I might do it again in 2020.

 

[newtekie1]

When the disk is rebuilding there is meant to be a huge amount of overhead slowing all access to your system or putting it out of commission for the duration of the rebuild.
Here's a (non-forum) article about the death of raid 6 in enterprises https://www.zdnet.com/article/why-raid-6-stops-working-in-2019/

The person writing that article doesn't seem to understand how RAID works or has never used one...

And URE doesn't stop the rebuild. In fact, modern controllers(and a lot of not so modern controllers) have options specifically to address errors on rebuild. All of mine by default just continue to rebuild. The URE will cause whatever data the read error happened on to be corrupt, but the entire array is not lost. And that's why you have backups. Any corrupt data can be restored from the backup.

As for huge overhead while rebuilding, yes there is. And yes the array is slower, but still usable and definitely not out of commission during the rebuild.

 

[Hellfire]

I let my NAS back itself up with a mirrored Raid.

If it is still within the NAS that's not a backup, that's a redundancy.

My backup system for my main data

NAS Drive for on site storage & back up of PC documents
- Google Drive for Backup1
- AWS for Backup2

90 day data retention on all files.

 

[Ahhzz]

If it is still within the NAS that's not a backup, that's a redundancy.

...

True, but considering at home, all I'm worried about is hard drive failure, it works as "backup" for me . I don't backup my main PC, as all I use it for is gaming, photo editing, browsing, nothing critical. The critical data (taxes, a very few docs) on my server is all double backed on the cloud. The photos are my main concern for "data retention", and the worst that could happen to me there is a drive failure. hence, redundancy equals one more level of backup for me

 

[Hellfire]

True, but considering at home, all I'm worried about is hard drive failure, it works as "backup" for me . I don't backup my main PC, as all I use it for is gaming, photo editing, browsing, nothing critical. The critical data (taxes, a very few docs) on my server is all double backed on the cloud. The photos are my main concern for "data retention", and the worst that could happen to me there is a drive failure. hence, redundancy equals one more level of backup for me

House could burn down.... that's worse than drive failure, but I get what your saying bro ;-)