Degraded Raid 10 Array

[doublejz]

I landed up hearing the dreaded system/BIOS beep and found that a RAID 10 array is degraded. When I go into the Adaptec Raid Configuration Utility I see that the first RAID 1 array #255 (within the Raid 10 array) is optimal while the second array #254 is degraded. If I attempt to rebuild it on the existing drive, it starts then errors out saying that there is an Inaccessible Disk (Dev02). I'm new to RAID 10 arrays and I believe this device houses all the educational and training videos and I really can't lose this data (no working backups since January). Looking at the below screenshots, should I just replace Dev02 with a new 2TB drive and attempt to rebuild it again?

Thanks in advance.

 

[ingtar33]

obviously one of the hard drives (dev-03) in the array is dead. you need to replace the failed drive in order to rebuild. rebuilding an array with a dead hard drive will fail every time.

 

[doublejz]

obviously one of the hard drives (dev-03) in the array is dead. you need to replace the failed drive in order to rebuild. rebuilding an array with a dead hard drive will fail every time.

Why did Dev02 become Inaccessible after trying to rebuild? I see that Dev03 was grayed out in the screen before trying to rebuild but I would thinkg Dev02 since its no longer listed would be the bad drive, right? Am I better off replacing both drives in the second array?

 

[molletts]

Dev02 is the one you need to replace. There's no need to replace Dev03.

It's worth pointing out, though, that none of the drives is RAID-rated, so overall array reliability may not be as good as with drives intended for use with RAID controllers. (It used to be possible to switch WD Blacks into RAID-friendly TLER mode using the WDTLER tool, making them a good cheap substitute for the WD RE drives that were intended for RAID usage, but that no longer works. Without Time Limited Error Recovery, the controller may kick drives out of the array unnecessarily.)

Also, remember that RAID is not a backup. You seriously need to get the backup working.

 

[ingtar33]

Dev02 is the one you need to replace. There's no need to replace Dev03.

you're right, i read that wrong.

 

[doublejz]

Yup, new replacement drive installed and raid rebuilt fine. I just need to see if they want to get actual raid drives. Theoretically, I should be able to get the same size raid drives IE. 2TB WD RE and swap 1 drive out at a time waiting for the array to rebuild on them until they are all replaced, right? Sorry, I'm a network guy but just trying to tackle other issues around the office too.

 

[molletts]

Yup, new replacement drive installed and raid rebuilt fine. I just need to see if they want to get actual raid drives. Theoretically, I should be able to get the same size raid drives IE. 2TB WD RE and swap 1 drive out at a time waiting for the array to rebuild on them until they are all replaced, right? Sorry, I'm a network guy but just trying to tackle other issues around the office too.

Yes, that's all that you'd need to do.

Make sure the backup is working first, though - there's always a slight risk involved in fiddling with a RAID array because you hammer it quite hard every time you rebuild it. I've never (yet) had a second disk die during a rebuild but I had a close shave once when a disk died the following day.

If there's space in the server, you could leave one of the desktop drives in and designate it as a hot spare. The controller will then automatically rebuild on it if a drive fails. If you enable copyback mode, it will copy the contents of the spare to the new drive when you replace the failed one (so the spare goes back to being a "reserve").

One thing to look at when buying drives for a large RAID setup is the non-recoverable/uncorrectable read error rate. This looks very low on paper (typically "less than 1 in 10¹⁴" for a desktop drive) but a rebuild requires a very large number of bits to be read across the surviving drives so, if you multiply it out, it doesn't sound quite as good. Your array requires about 4×10¹³ bits to be read during a rebuild (4TB; I usually work on the basis of 10 bits per byte to allow for overhead, etc. so 40Tb). WD RE drives have an unrecoverable bit error rate of <1 in 10¹⁵ (likewise Seagate's Enterprise Capacity range) which gives you that extra little bit of comfort.

Hope this is useful
Stephen