Bad Stripe

[curran]

Should I be conconcerned that I have a couple of servers stating they have a bad stripe on the arry??? From my understanging this like an indicaton of a bad segment on the disk?? Any help would be appreciated

 

[elconomeno]

an IBM technician came by last week.
He told me the following

if you have bad stripes, you have to delete the array and create it again and install a fresh OS

I dont have experience with bad stripes, because it was our lucky day, we didnt have bad stripes

 

[scsiraid]

A bad stripe typically occurs when you have been running a RAID 5 array in critical mode (1 drive bad) for an extended period. If an unrecoverable read error occurs in this situation... you cannot recover the data since the array has already lost a drive and cannot deal with losing another one. When this happens... the controller can do one of two things... it can flag the stripe as bad and 'lose' the data in the stripe, or kill the whole array. ServeRAID goes the bad stripe route.

 

[technome]

scsiraid..

Glad you brought this up.
Most raid 5 users do not know block loss, in degraded mode, increases the danger of a raid 5 array loss. With increasing drive capacity/drive count in an array set, the danger increases.

The linked article from Intel, is about raid 6, but refers to the issue you mention with raid 5 block loss.

http://www.intel.com/technology/magazine/computing/RAID-6-0505.htm

........................................
Chernobyl disaster..a must see pictorial

http://www.kiddofspeed.com/default.htm

 

[scsiraid]

Yup.. Good Article. I know one of the authors. Good guy. One terminology difference between this article and ServeRAID speak is that what Intel refers to as a 'strip' is referred to by ServeRAID as a 'Stripe Unit'. The 'poopoo'ing of 2dxor is probably a bit harsh. RAID 6 with 2dxor isnt as bad as they say. We have seen about 10% impact over RAID 5. ServeRAID 8i supports RAID 6.

 

[technome]

One terminology difference between this article and ServeRAID speak is that what Intel refers to as a 'strip' is referred to by ServeRAID as a 'Stripe Unit'"

I get a kick out of IBM and Microsoft... what ever terminology term is used, they can not resist creating a new name for it...actually a good marketing ploy.

Wish we had a coprocessor war ( major coprocessor speed increases), so parity calculation overhead is not a main concern. I spend too much time compensating for parity overhead.

Looks like raid 6 or an equivalent is becoming more important with the increasing array sizes, raid 5 is beyond it's safety limit with the larger arrays.

........................................
Chernobyl disaster..a must see pictorial

http://www.kiddofspeed.com/default.htm

 

[scsiraid]

Another major factor pushing RAID 6 is SATA's push into the enterprise. The lower MTBF and larger capacities for SATA greatly increases the probability of a double drive failure prior to rebuild completion.

 

[technome]

Have not used SATA for a production server yet but I would be tempted to skip raid 5 or 6 and jump to raid 50 or raid 10 with a hotspare, considering the drive costs are lower..love to try out the Tekram ARC 1120 or 1160, as expensive as they are.

........................................
Chernobyl disaster..a must see pictorial

http://www.kiddofspeed.com/default.htm