Take the 2-minute tour ×
Server Fault is a question and answer site for professional system and network administrators. It's 100% free, no registration required.

I'm trying to set up an offsite, offline backup procedure for my company's critical information. We already have an in-office redundancy backup set up using RSYNC that backs everything up onto two separate hard drives once a day, and we would like to take our existing backup system and duplicate it onto a third drive that we can then remove and carry offsite every month or so.

My question deals with the duplication step. We're looking for something simple but automatic (or easily able to be automated). My initial thought was to recast one of our current backup drives as a RAID1 pair, get a hot-swap drive bay (that is, something like this Antec Easy SATA http://store.antec.com/Product/accessories-other/easy-sata/0-761345-30750-5.aspx), buy a few extra drives, and swap them in as needed, letting RAID rebuild the drives automatically as we go.

The problem is, I have no experience working with RAID, so I don't know what to expect. I'm trying not to re-invent the wheel. In my mind, using RAID is ideal because

  • The addition of a new backup drive would be transparent to our current setup, and not require any changes outside actually setting up RAID
  • The added procedure for backing up the data would be reduced to
    1. Grab the current drive
    2. Hot-swap in a blank replacement drive (or a previous backup to be overwritten)
  • The system will automatically treat the replacement drive as a failure and rebuild it; when it is done rebuilding, the drive will always be kept current and ready to be removed

Does RAID actually work the way I'm thinking? Or is there an easy way to make it behave this way? If not, is there another setup that will be this simple to maintain, just grab and go?

share|improve this question
    
How much data? For small amounts (under 1TB) you can use an RDX drive for your offsite swaps. RAID was not meant to be used this way... –  TheCleaner Feb 22 '13 at 22:19
2  
Yeah, this is a janky approach to backups. –  ewwhite Feb 22 '13 at 22:29

5 Answers 5

From my point this is a very bad idea, RAID 1 was never designed to be a backup solution, but a redundancy tool. That said there are tons of tools that allow you to backup a complete drive ( snapshot) which will work rather fast as well, for instance drive image XML on windows. Linux certainly has the same or similar tools available.

share|improve this answer

Well, from a high level, this sounds plausible. But, as they say, the devil is in the details. Who would be swapping the drive? Will you ALWAYS verify that the other drive hasn't failed before pulling the backup drive?

A backup isn't a backup without a restore procedure. What is yours? 5 years from now will you be able to purchase a RAID card that will read this drive?

My suggestion is to find another way.

share|improve this answer
    
+1 for mentioning the raid card.. I hadn't even thought about that (assume they're using linux software raid). But true. What if they plug the drives in in the wrong order (is this still a thing?)- mdraid used to get really unhappy if things weren't where it expected them to be. –  Tom O'Connor Feb 23 '13 at 8:32

At a very basic level, yes this works. The controller will consider a newly added drive to be failed and will mirror the working drive's contents onto it. The harder question actually comes in maintaining a consistent state for the backup drive upon removal. You'd be challenged to find a way to ensure that the filesystem was in a consistent state upon removal of the drive unless the system is shutdown during removal, though.

With that said, I think you're likely better using a higher-level solution that is aware of the filesystem for synchronization. That will allow you to flush out writes to the drive and be point-in-time consistent.

I think that the overall preference I'd have for something like this is along the lines of RAID 1 < rsync < (http://www.bacula.org/en/, http://en.wikipedia.org/wiki/Bacula)

share|improve this answer

Since you mentioned RSYNC, I'm going to assume your servers are Linux based.

I would recommend using a hot-swappable eSATA drive or external USB drive. Then, script RSYNC into a cron job. Note: eSATA throughput is a lot higher than USB throughput, so RSYNC will finish faster on eSATA, but if your system doesn't have eSATA, you would need to invest in an expansion card for it.

You could also do the same with a portable NAS.

share|improve this answer

I have successfully accomplished offsite backups using raid 1. I suggest you use mdadm under linux for this process. mdadm is idea because no matter how many years down the road you will always be able to mount and retrieve your data under linux. Build a raid1 set just like you said. Each month mark the drives (use a sticker on the physical drive) and the date that it was inserted or removed. This will guarantee that you always pull the older drive. Before pulling the drive you should do these steps

  • Check the status of the array with mdadm --detail /dev/md0
  • force the system to re-verify parity of your data with echo "check" > /sys/block/md0/md/sync_action
  • shutdown the machine, pull the drive, put in the new one and add it back into the set

If you can't have your data go down, you can mount the drive in readonly mode for a period and then hot yank the drive. If you are using sata drives then you can add and remove them while the system is running. Here is a link with more of the specific commands:

http://www.howtoforge.com/replacing_hard_disks_in_a_raid1_array

share|improve this answer
2  
RAID is NOT a BACKUP. –  Tom O'Connor Feb 22 '13 at 23:25
    
You are correct RAID is not a backup. RAID only protects from hardware failure. If all of your files get deleted, or you get a virus, the RAID array will happily duplicate these bad things across all drives. To truly make backups, you need to make separate copies of the files which are dated and no longer get written to. The asker would like to maintain an offsite back (after all a fire in the building could destroy ALL copies). He wants to use RAID1 as a simple method for duplicating the data. I am simply helping him on his way. But thanks for your enthusiasm at least. –  portforwardpodcast Feb 23 '13 at 2:48
    
I simply don't think it'd work without massive risk being incurred when you unplug the disks. Also, what if you put them in in the wrong order, or something goes wrong. I just wouldn't do it. –  Tom O'Connor Feb 23 '13 at 8:30
    
RAID1 has no order. Also failing drives out of mdadm is 100% safe. I honestly think that this is a good solution. –  portforwardpodcast Feb 23 '13 at 9:29
    
OK, but I won't sanction it. –  Tom O'Connor Feb 23 '13 at 11:22

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.