My Experiences with Hardware Raid

[tojoski]

So I thought I would share my experiences with running a hardware raid in my home server.

 

My first home server, well the first I had that was really designed with bulk storage in mind, was based on Windows Server 2008R2, with a

small 80Gig Sata drive as the OS drive, and (5) Western Digital 1TB (WD10EADS) storage drives.

 

The raid controller was an Intel SRCSAS18E, picked up from ebay for $100 and a battery backup unit (BBU), picked up from ebay

for about $40. SAS cables also picked up on ebay for about $20.

 

I picked the WD10EADS drives specifically because the supported a feature called "TLER" or "Time Limited Error Recovery".

The 2TB green drives of this era (WD20EADS) also supported this feature.

 

Desktop class hard drives will go into a "deep recovery" of sorts if they run into a bad sector, where the drive stops all data transfer and

tries for an indeterminate amount of time to recover. The problem is that after about 10 seconds, most raid controllers consider the drive to

have failed, and drops it from the array.

 

So with TLER on the WD drives, with a dos executable from a bootable flash drive, you can set the maximum amount of time the drives will enter

this state. The default when you enable TLER is 7 seconds.

 

My array rocked along for about 6 months, and one evening I was alerted to a degraded array by a very loud warning beep. (this is a nice feature

of hardware raid controllers)

 

I requested an advanced RMA from WD, got the new drive in a few days later, enabled TLER on the new drive and put it in the server.

Two days later the array was finished rebuilding and all was well..until about 8 months later

 

The same thing happened, although this time when I got the new drive in, the TLER functionality had been removed!

 

Turns out, TLER is a feature that WD specifically put into their RE (raid edition)series drives, the fact that it was present at all in the

EADS series green drives apparently was unintentional.

 

So after fighting with WD for approximately 2 months, meanwhile my array is in a degraded state, they wouldn't answer why TLER was removed from these

mid product-cycle.

 

Eventually, WD agreed to replace all 5 green drives with new RE drives.

 

Recently, I've heard of people getting these EADS drives and being able to enable TLER, so maybe WD put this functionality back?

 

I'm not sure, but I can say, if you are planning on building a raid array, check out the list of supported drives for that specific raid controller.

 

Sorry to be so long winded, but its a long story.

[pcdoc]

Great write up and thanks for the feedback. Was not aware WD had done that with EADS drives. I have a shelf load of these things so I should check mine. Glad you got it resolved.

[DutchWHS]

I'd heard about RAID, but did not have any hands-on experience with it. For me this was an excellent post, I now see the importance of TLER. I'm starting to understand why people are not so keen on using RAID ... there's just so much you have to consider!

 

Thanks for your writeup!

[usacomp2k3]

It took 2 days to rebuild the array? How much data was that? What would have happened if another drive had died during that time?

[tojoski]

It took 2 days to rebuild the array? How much data was that? What would have happened if another drive had died during that time?

 

It was something like 3.6TB, but only about 600GB of data.

The rebuild time is based on the size of the array, and not the amount of data thats on it.

 

With RAID5, you're only protected against a single drive failure, so if another one fails during rebuild, your array and your data are hosed.

 

This is why RAID6 is becoming more popular, as it protects you against 2 drive failures, but of course you loose 2 drives worth of capacity.

 

Its important to remember though that any form of RAID is not a backup, and any data that is important to you should be backed up.. or duplicated

 

Here's a good article on RAID5, worth a read: http://www.zdnet.com/blog/storage/why-raid-5-stops-working-in-2009/162

 

Great write up and thanks for the feedback. Was not aware WD had done that with EADS drives. I have a shelf load of these things so I should check mine. Glad you got it resolved.

 

You need the WDTLER utility. If you can't find it, shoot me a PM.

[tojoski]

BTW, I forgot to mention, if you want to run RAID5, thats a nice card, and you can find it cheaper on ebay under its LSI counterpart: LSI 8408E (its the exact same card)

Edited December 17, 2010 by tojoski