HP Smary Array Rebuild Problem

[freedomit]

So you might remember my post about a possibly failing HDD in a HP Gen8 Microserver..
https://www.technibble.com/forums/threads/hdd-failing.77128/#post-621153

I decided to order an advanced RMA from WD and replace the drive. So i powered down the Server (they are non hot swap), swapped the drives and rebooted. The server detected the drive was missing and booted into Server 2012. The problem i have is that i cant get it to rebuild. When i open the Smart Array Managed it just complains about the failed drive and gives no option to rebuild. I tried powering down the Server and booting with one drive, then shutting down and plugging in the new drive but still the same.

Also, when i boot i get the following error message...

1787-Drive Array Operating in Interim Rcovery Mode
The following disk drives are insufficient size and should be replaced - Bay 1

The drive is a WD Certified Refurb but its the same size, model number, block size and LBA so why am i getting this error message.

Anyone got any ideas? I wish i never bothered replacing the drive now!

 

[freedomit]

So i have done some more digging into the HP Array Diagnostic report and found this...



So it looks like they are actually different sizes

@lcoughey maybe you can shed some light for me?

 

[YeOldeStonecat]

Yeah RAID won't rebuild if the "new" drive is smaller...need at least equal size.

 

[freedomit]

Yeah RAID won't rebuild if the "new" drive is smaller...need at least equal size.

Thanks for confirming thats definetly the issue. I just dont understand as its an RMA replacment from WD, same model number and everything so why is the size different? I guess this is why its best not to use 3rd party drives and stick with the slower & expensive HP options.

Could it be that WD have marked some of the blocks as bad and retired them then recertified the drive as good?

 

[lcoughey]

Could be that there is an HPA set on the replacement drive. But, I'd just return it back to Western Digital and tell them to send a working drive with the full sector count.

 

[lcoughey]

...one other thing. As I suggested before, you should be putting a new drive into the system, not a crappy refurbished drive. Your client's data and your time surely has to be worth more than the cost of a new hard drive.

 

[freedomit]

...one other thing. As I suggested before, you should be putting a new drive into the system, not a crappy refurbished drive. Your client's data and your time surely has to be worth more than the cost of a new hard drive.

I wish I could, WD no longer make this model and the only ones I can source from distribution/resellers are refurbished. I thought going down the RMA process would get me a new drive but I was sent a refurb.

The only other option is to replace both drives in the Raid 1 but it’s not exactly a simple process.

 

[lcoughey]

The only other option is to replace both drives in the Raid 1 but it’s not exactly a simple process.

Why isn't it? RAID 1 is just a mirror, which means the data should be directly accessible from either drive when connected directly to a system, even if you do have to load the file system with R-Studio.

Put new drives in the unit, setup a new RAID and then saved data from original drive(s) back to the newly setup array.

 

[freedomit]

Why isn't it? RAID 1 is just a mirror, which means the data should be directly accessible from either drive when connected directly to a system, even if you do have to load the file system with R-Studio.

Put new drives in the unit, setup a new RAID and then saved data from original drive(s) back to the newly setup array.

Your an expert in this area so I’m sure your right but I would always be nervous about cloning drives in a HP Smart Array. I know the meta data for the raid is on the drive and I don’t understand how well that transfers when cloned to other drives. Also, sometimes you need Smart Array Licences to shrink/expand arrays.

Other problem is downtime, shutting down a server and swapping out a drive is easy to schedule. Trying to schedule an hour or two downtime for a clone is harder, and that’s if we do the job onsite.

If we did go down this route would it not be better to install a new drive, rebuild, pull the other drive and insert new drive 2 and rebuild? A risky process but should work.

 

[DataMedics]

The only other option is to replace both drives in the Raid 1 but it’s not exactly a simple process.

It could be a very simple process. Sell your client on an "upgrade" to 1Tb or 2Tb drives for future expansion. Put in one 1Tb drive, let it rebuild onto that, then put in the second one and let it rebuild again. Then you'd just have to expand your virtual disk in the Smart Array console and adjust your partition size.

 

[freedomit]

Ok so i spoke with WD support...apparently what happend is that WD did not have an exact replacement available so they sent a different product number but same model number. Normal procedure is to send the customer an email to confirm if this is ok but they admitted in this circumstance they didnt and just sent it. They are going to send an exact product number drive to me now but it might be a few days depending on availability.

 

[fencepost]

Some years back I actually started underprovisioning a little bit on RAID setups on Dell PERC6 controllers, leaving a few MB unallocated on each drive when creating volumes. I actually haven't had to deal with a replacement where that was relevant, but my thinking was that if I couldn't get an exact model as long as there was enough space for the RAID volume I should be OK - and the cost of a few MB of capacity was well worth it.

 

[freedomit]

UPDATE:
WD support have come back to say they no longer make the drive or have spares so they can only offer the replacment already sent or a refund. Im going to take a refund and then look at swapping out both drives to build a new raid.

 

[Sky-Knight]

You don't have to worry about cloning the RAID meta, that stuff is in the partition table but the OS doing the cloning work shouldn't be able to screw that up once the controller is online.

Honestly, I'd get two new 1tb drives, image the 500gb to one of the 1TB disks, boot the system on the new 1tb disk, then engage the mirror.

Heck the details you've posted indicate it's just a SATA drive, you could probably get away with a pair of 512gb Samsung Evo Pros.