Backup file server archetecture question

[d259]

We are looking to upgrade our "backup" file server at work (small business). Current setup is a Ubuntu box with a 160GB Samba share that is backed up periodically by our secretary onto a USB Passport drive.

Server requirements: redundancy on the file server and off-site backup (no online backup tools are allowed due to the nature of our business). We don't have a huge amount of data to store (~100GB) but could change. Must be simple as I'm almost positive our secretary will continue to take care of the backups

Our main IT guy and I have differing ideas on how to implement this (he doesn't have any experience with RAID arrays whereas I have several years of managing my own 2+TB servers) -

His idea:
Hot-swappable hardware RAID1 setup, two drives in that one that is pulled weekly for someone to take home (off-site backup) and a new drive replaced to do a rebuild of the array. Hard drives are rotated through periodically so you would have approximately a month's worth of archive as well as backups.

My idea:
Any type of RAID in server that does a cron rsync every night of the data to an external hard drive. External hard drive is swapped out periodically for a new one for archival purposes.


I already had an hour discussion trying to convince him that having a RAID array do a forced rebuild every week would seriously compromise the integrity of the drives, to no avail. I'm looking for some guidance/experiences that I can take to him and try to steer him away from his idea.

Any suggestions?

 

[sondo2121]

Our main IT guy...

hah! main....you being the other one?

My Thoughts:

• Continue to have a server at your workplace
• Have a RAID array on it (1, 10, 0+1, 5, etc...)
• Have 2 external hard drive where the secretary would plug in and use script or executable of some sort to sync (rsync) the files from the RAID to the external.
• Have her rotate the external hard drives from her house (or somewhere else) and bingo, off site backup

My thought: 'yourway': The reason why I suggest this is because I know fabs and the situation he's in...there is more than one way to do this though....

 

[keenan]

Well, most of what I've seen suggests that drive failure isn't really correlated, but I still like your idea much better for several reasons. First, while rebuilding the array is totally vulnerable, and a disk failure will be a bit of a mess to clean up. Second, I have seen controller bugs that lead to the wrong disk being selected as master in RAID1; probably rare, but rebuilding can be a bit of an error prone process and should be avoided, certainly not intentionally used as part of SOP. Third, even hot-swap cages aren't really designed for repeated insertion/removal, and the disks aren't designed to be carted around every night either. I wouldn't be surprised if the hardware wore out pretty quickly and you started getting flaky connections and such, which isn't fun to deal with in a RAID situation. Finally, yanking a running image of a live filesystem could very well leave you with plenty of data loss in your 'backup' if the right conditions are met when the disk is pulled.

Your solution is much better. Dedicated backup media, filesystem-level reads, redundancy retained on the active media...

 

[phildabill]

As you stated, doing a rebuild of the data every week could compromise the disk, but is also redundant. It seems he does not understand the nature of a rebuild. Both drives will be stressed for no reason. Make sure he understand what a rebuild entails.

Explain to him that rsync, unlike a rebuild, only transfers files whose timestamps have changed (if you do a CRC with rsync, it will end up being almost the same as a rebuild).

Does the off-site have an internet connection? If so, do the first rsync locally, then leave the external drive off-site and update it every day. This will remove a step from the process (moving external drive).

Another option would be to try both for a predetermined period. This will demonstrate how your method is less time consuming.

 

[d259]

Does the off-site have an internet connection? If so, do the first rsync locally, then leave the external drive off-site and update it every day. This will remove a step from the process (moving external drive).

I did pose this to my boss but he is extremely security conscious and like I said before the nature of our business won't allow all that data being transferred to an off-site location. All the data must be under physical control.

 

[brandinb]

I would set the data on a 2 disk raid 1 on the server. I would install another drive (permanently) in the server ad a dedicated local backup (it sucks retrieving backup media from offsite to recover) and script the backups to go there. I would then use a tape drive or usb hard drive additionally to take the backup off site. Perhaps explain to your boss high level encryption techniques available via vpn now days. It works for the government and medical institutions so why not your data? I would also consider keeping multiple copies of backups locally as I have experience recovering data from backup and it is nice having multiple recovery data sets just in case. As for a usb disk I would buy a nice ssd drive and put it in a enclosure that way if its dropped the data is not "gone". heh I would rotate out 2 or 3 drives for remote backups aswell.

 

[k1pp3r]

What is your budget, you may get off cheaper buying a solution rather than building your own

 

[outspoken]

budget is key. raid is a must.

some friendly advice: don't go outside your departmental duties or over the heads of others in small businesses.

 

[d259]

What is your budget, you may get off cheaper buying a solution rather than building your own

I haven't heard specifics yet but probably no more than $1-2k. I just rebuilt my server (5TB storage) for $700 so I can't fancy going higher than $1k.

some friendly advice: don't go outside your departmental duties or over the heads of others in small businesses.

Not to worry, I'm not trying to go over anyone's head and certainly not trying to discredit the IT guy. I took the first few years of CCNA classes and from that experience I deliberately steered my career path towards engineering and away from IT. I'm just the kind of person that when I see something that has the potential to go horribly wrong (especially with data storage) I have to speak up. There are just to many people in the company that work directly off the data that is stored on the server to take chances so I'm trying to steer him away from what I perceive to be an iceburg of problems ahead.