Raid5 - Missing Member - yet Disk Healthy? 1

[schase]

Howdy folks.

On my PowerEdge 2500 server (Raid5 - 3 disk config) I have Disk0 showing degraded in Array control, (both in startup and in the Dell Open Manage) yet in the Disk Manager view it says all three disks are healthy, and green blinking lights on all three disks.

I do have the error warning indicator going off though, and notices saying redundancy lost.

So - what to do? Earlier one of the disks was blinking yellow (but green now).


Stuart

 

[kmcferrin]

Could have been a transient error. The other thing that I've seen on those Dells is that the disks can vibrate loose from their connectors over time. You should be able to just unplug the problematic disk, wait 30 seconds, then firmly plug it back in. But only do this for the disk that is generating alerts. If you pull the disk then your server will see it as failed, and when you plug it in again the server will probably try to rebuild the "failed disk" on the disk that you just plugged in (operating under the assumption that it's a new disk). Pulling a second disk while the array is rebuilding with cause a failure with data loss...you'll have to restore from a backup at that point.

 

[schase]

it looks like you may be right. I have not yanked Drive 0 yet, but now I'm getting memory errors.

Going to play further and see whats up.

Stuart

 

[schase]

pretty much spot on,

Ran a check on the memory - cleared that error. Then went into Raid Configuration Bios and with directions from Dell, You get to the manage containers section and you see Status of Critical - with one missing member. Then yank the offending harddrive, waited the 30 seconds or so you mentioned and replaced it. Status went to rebuilding

And a little more than 2 hours later that drive finished rebuilding, rebooted normally.

Stuart

 

[kmcferrin]

Good. Just keep an eye on the drive...it could certainly be a warning sign of things to come, but if it's working and the array is already redundant, then I wouldn't worry to much unless you get a predictive failure warning.