0

Given: 2 HDD in RAID and case when we lose 1 of them after 3 years:

  • If we use RAID 0, we lost data
  • If we use RAID 1, we have a copy on possible also bad-health disk, any restore may cause failure also
  • If we use RAID 5, same as RAID 1, but we can lose another one disk and all the data during a rebuild
  • If we use RAID 10, same as RAID 1, all disks work pretty the same time and have a pretty same worn level.

So, what is the benefit of using RAIDs, if we can simply configure every night backup, e.g. only for diff of data? It such a situation we can lose data only for a day (if it if appropriate, e.g. a home file server), but the backup disk will have so reduced worn level comparing to the main disk.

Maybe I am wrong What do you think?

alexloiko
  • 3
  • 1
  • 4
    Does this answer your question? [Why is RAID not a backup?](https://serverfault.com/questions/2888/why-is-raid-not-a-backup) – Lenniey Feb 26 '20 at 11:56
  • Do you want to reinstall your operating system and applications, reconfigure the system, restore your data, and verify that everything works as it did before? How much time would that take? How much would it cost you in lost time, productivity, and business revenue? Not having to do that because of a disk failure are just a few benefits of using RAID. – joeqwerty Feb 26 '20 at 11:57
  • 2
    Your assumption is wrong. RAID5 need minimum of 3 disks, RAID10 need minimum of 4 disks! – Romeo Ninov Feb 26 '20 at 12:00
  • But how to deal with a situation, when after failing of one disc in RAID, all other discs are in the same pre-failure situation. You can lose all the data during trying to recover. – alexloiko Feb 26 '20 at 12:17
  • 1
    Yes, this has in fact happened to me and is a sanitary lesson - RAID 5 is flawed in this regard. RAID 6 is better, RAID 10 is better yet, but there is *always* a risk. You should do *both* backups *and* have redundant storage, and even then you're not guaranteed not to have an outage as restoring backups to a new machine can be much less simple than envisaged. – Alex Berry Feb 26 '20 at 13:27
  • 1
    Repeat after me: *[RAID is not a backup, RAID is not a backup, RAID is not a backup, RAID is not a backup, RAID is not a backup, RAID is not a backup, RAID is not a backup ...](https://serverfault.com/questions/2888/why-is-raid-not-a-backup?noredirect=1&lq=1)* – shodanshok Feb 26 '20 at 14:28

4 Answers4

5

RAID is not for backups-it's for availability. If you lose a disk in a RAID setup, your systems can still run. Without RAID, you would have to buy and prepare new disks, and restore your backup before your business can continue.

Bert
  • 2,733
  • 11
  • 12
1

Raid 5 is not same as raid 1, raid 5 is striping with distributed parity. For a raid 5 you need to have a minimum of 3 disks atleast.

Different types of raid not only protect your data but also provide read/write performance. RAID is not a backup, its data protection in case of a failed drive.

user561690
  • 21
  • 4
  • I understand that raid is not a backup. But in case of failure of a disk in RAID, there is a big chance to lose some another disc in RAID and all the data. So, if we can lose all the data in any case, why to use RAID? – alexloiko Feb 26 '20 at 12:14
  • Anything is possible, but it's not a given, and it's unlikely, that you'll experience multiple, simultaneous disk failures. Using that as a reason against using RAID is ill-informed and shortsighted. – joeqwerty Feb 26 '20 at 12:38
  • right said joeqwerty. @alexloiko, Raid is redundancy, imagine a data center running a JBOD and disk fails, you will not like to wait for the backup to finish and then start your operations. – user561690 Feb 26 '20 at 12:46
1

As Bert said, RAID is for availability and/or performance, backup is for disaster recovery. Backup (and snapshots) also help against accidental data loss by accident or mistake, which RAID does not. Backup and RAID simply are not on the same spectrum. With modern file systems like ZFS, the areas are somewhat overlapping, but backup is still required.

Before we set up a server, we should evaluate the importance of the data protection, the required performance and the cost of potential downtime. This gives us the frame to plan the RAID level and so on.

Now RAIDs 1 and up can prevent unwanted downtime. As OP says, any rebuild poses some risk of finding another failed disk. But with the RAID running degraded, we have the time to plan ahead: Make plan A and plan B, assign a downtime, prepare the replacement parts, check if the backup is current and so on. Maybe we can test-run the procedure on our identical training server / organ donor. You have one, right? Because it is important.

And data is important, so we must act carefully and precisely. RAID buys us the time to prepare. It is even possible to just buy a new server and migrate next weekend without risking the rebuild.

In my case, with about 50 Petabyte disk space backed up to a centralized tape systems, disaster recovery can take several days. The incremental backup runs about 12 hours against the 50M files.

RAID 10 in this case protects against downtime, and the data loss between last backup and current status. But usually I try to replace the server before the disks reach critical age.

RAID allows us to act professionally, not react hurriedly.

Posipiet
  • 1,725
  • 14
  • 13
-1

Best RAID configuration might be indeed no RAID at all: https://blog.shi.com/hardware/best-raid-configuration-no-raid-configuration/

Besides that, and at least for some years now, RAID 5 is officially not recommended at all by Dell for any business critical data: https://eqlsupport.dell.com/WorkArea/DownloadAsset.aspx?id=6442454665 https://community.spiceworks.com/topic/251735-new-raid-level-recommendations-from-dell

Raulinbonn
  • 54
  • 5