Food for thoughts - RAID / Backup

[Marie.O]

Today, during my surfing threw the web I came across an interesting tid bit.

A fairly high class installation of a networked media system in a house. They've invested a few 10k EUR I can imagine. And they did not go RAID.

The stated reason was, running all disks the whole time stresses all disks identical. This results in the same fatigue for all disks, and quite probable make them fail at the same time.

Instead, they opted for a simple NAS doing backups at night. During the day, the HDD in the NAS are kept powered down.

With todays prices for storage, this might be a worthwhile idea for LinuxMCE users as well.

[Zaerc]

Guess I'm just going to have to keep repeating this until hopefully someday it sinks in...

Using RAID is no alternative for making backups!

[Marie.O]

Zaerc,

you are 100% correct.

And my post merely highlights another reason for it. Lots of people mistake RAID for backup, where in fact they achieve slightly longer uptime, where they could go out and look for a replacement drive.

[b4rney]

Agreed. I lost my raid array moving from 0710 to 0810.

I now have a dedicated NAS backup (not mounted by lmce) so it isn't spinning 24/7. Still using software raid but thinking of reverting to single disks. My raid disks have been spinning for months without a rest.

Best solution IMHO would be a rsync incremental overnight backup of the lmce media drive(s) to the NAS. Just need a NAS that supports rsync.
Barney

[valent]

Agreed. I lost my raid array moving from 0710 to 0810.

I now have a dedicated NAS backup (not mounted by lmce) so it isn't spinning 24/7. Still using software raid but thinking of reverting to single disks. My raid disks have been spinning for months without a rest.

Best solution IMHO would be a rsync incremental overnight backup of the lmce media drive(s) to the NAS. Just need a NAS that supports rsync.
Barney

What did you do with your raid while moving from 0710 to 0810? I have moved mine and didn't lost a bit of data. I also have LVM and LUKS encryption on top of RAID-5.

Also added a new disk recently and while growing raid-5 field power went off! I thought that data was gone, but after consulting with few linux gurus I got RAID to rebuild and grow without any issues.

@posde The wear on disk is never the same, here probability and chaos theory comes in place. It is statistically almost impossible for all drives to fail at same hour or day, let alone AT THE SAME TIME. Complete PR bullshit if you ask me.

My strategy is to change drives every 3 years with new ones, no matter if they had a failure or not.

[Marie.O]

@posde The wear on disk is never the same, here probability and chaos theory comes in place. It is statistically almost impossible for all drives to fail at same hour or day, let alone AT THE SAME TIME. Complete PR bullshit if you ask me.

valentt, a customer of mine had 3 Cheetah disks fail within a one week period.

[b4rney]

Valent,
I attempted to install lmce 0810 and it detected the raid 5 array.
The array was one disk down and showed as rebuilding. This never completed though.
I had problems getting lmce 0810 to work so reverted back to my 0710 build on a separate hdd.
I think it was the reversion to an earlier version that bricked the array.

Now I backup my backups. The tough lessons are the best!
Barney

[valent]

@posde The wear on disk is never the same, here probability and chaos theory comes in place. It is statistically almost impossible for all drives to fail at same hour or day, let alone AT THE SAME TIME. Complete PR bullshit if you ask me.

valentt, a customer of mine had 3 Cheetah disks fail within a one week period.

That is why I check my raid once per week, but I'll write a script to send email and sms alert if/when any hdd has errors. Week is plenty of time to replace hard drives. Then it was obvious you got drives from a bad batch and they all were bad from start, or there was some other issue causing drives to burn out, like a bad PSU.

Choose which is worse:
1. HDD dies and you are left without data on it.
2. HDD dies and you have one week to replace it so that RAID continues working, and no data is lost that was on that drive.