RAID failure questions

Admiral Huddy · 9 Mar 2009 at 09:46

I have a RAID-5 setup in which one of the disks shows as an error. Therefore, as a result, the RAID has degraded.

I have removed the offending drive, and plugged this into a USB caddy on another system for testing.. I ran HD tune disk checking (full check) and it reported no errors?

How can this be?

Also, I can get the same Samsung 500Gb 16MB cache drives (these where just before the F1s), so will any 500GB drive be okay for a replacement.

rpstewart · 9 Mar 2009 at 09:55

RAID arrays can be a bit paranoid when it comes to marking drives as dead. Quite often the cause is the drive failing to respond quickly enough, possibly because it's been busy remapping a dodgy segment (hence why HDTune came back with no errors).

Before you go down the road of purchasing a replacement I'd run the Samsung diagnostics on the suspect drive, if a full scan with those comes back clean then you could stick the drive back into the array and hope it's just a glitch.

In theory any 500Gb drive can be used as a replacement although I'm not sure how the array will react if there is a slight size difference in the disks. Bigger is fine, it's what happens if the new drive is too small that I'm unclear about.

Admiral Huddy · 9 Mar 2009 at 10:23

Cheeers RP.. I knew you would by in the know.

This was my thinking.. The RAID-5 does appear to be a bit tempermental.

Admiral Huddy · 16 Mar 2009 at 11:58

Ok, I replaced the drive that failed, and left the PC on for a few days whilst it rebuilt the array.. Everything works ok. However, the PC appears to be lagging and isn't as quick as it used to be. It's only aboiut 10% slower but this is more noticeable in Photoshop.

Any thoughts?

JonJ678 · 16 Mar 2009 at 20:26

Are you certain it has finished rebuilding? Beyond this, what else have you changed and where does the 10% slower figure come from? Does the new drive differ from the old?

I'm quite surprised that you replaced the drive so quickly, Id have zero'd it and put it back in to see if it rebuilt happily. Depending on what system you're running the raid 5 on, you could now zero and add in the 'broken' drive. More capacity always good.

Something like: http://isomerica.net/archives/2007/07/12/growing-a-linux-software-raid-5-completely-online/

Admiral Huddy · 17 Mar 2009 at 12:13

def finished rebuilding - I checked the RAID menu and everything shows as normal.

I couldn't get the same drive but replaced it with a drive of same spec (ie. speed, cache, capacity, latency etc) despite it being a different make.

inph · 17 Mar 2009 at 12:43

JonJ678 said:
I'm quite surprised that you replaced the drive so quickly, Id have zero'd it and put it back in to see if it rebuilt happily.

I would have started with this after the error check came back clean.