Do you spread hard drive purchases amongst different sellers?

[Deleted47050]

I know this is considered a good rule of thumb to minimize the chance of stumbling upon a faulty batch of drives, however I was curious to see how many of you did that?

I am currently in the market for 4 1TB WD green drives for a small FreeNAS box running RAIDZ2 so this is something I am considering. What is holding me back is "if the drive model is the same, how can I know the drives are not from the same batch anyway"? I mean, is this so hit and miss?

 

[depasseg]

I don't bother worrying about it. I just buy, run the burn-in and then use them. Once you accept that they will all fail, and plan for that failure, decisions like this are easier.

 

[zambanini]

does not make any sense ( to take care about the batch numbers) when you use 'green' disks.

 

[Deleted47050]

does not make any sense ( to take care about the batch numbers) when you use 'green' disks.

Why is that?

 

[joeschmuck]

Here is the way I look at it... I do not worry about batches so much because all my important data is placed on DVD periodically so that my NAS is not my only backup. If I lose a backup of my main computer, that is fine because I'm banking my main computer doesn't fail at the same time as my NAS. Also, odds of an entire batch of drives failing at the same time are slim. When I do have a drive failure, even though I have a RAIDZ2 setup, I'll shut down my NAS until a replacement drive arrives and I can resilver and new one into the system as to minimize potential data loss like my media collection.

If you do buy drives from the same lot, this could be a good thing when the drives are beyond warranty and if you have a few failures, say the motor fails in one drive and the controller fails in another, you can replace the failed controller and be up and running again with that drive. I'm not saying everyone would do that but some folks do, especially if they have valuable data and they need to recover that drive at all costs. And if the motor fails on the drive which has the data you need, you can relocate the platters but it's not for the faint of heart. I've done it once for the hell of it and honestly didn't expect it to work but it did. I got lucky. I'm not positive I'd be so lucky with today's super high density drives.

 

[Fuganater]

Nope. They are either good drives or bad ones. If they fail the burn-in test than return it and get another.

This is coming from a super impatient person though.

 

[danb35]

They are either good drives or bad ones.

It really isn't quite that simple. I don't spread out my purchases, at least not intentionally for this purpose, but there is some logic to it. Certainly, if a disk is dead or defective out of the box, burn-in testing should determine that. But there are more subtle defects that could cause drives to fail early, or even cause lots of drives to fail right about the same lifetime (which would suck if they comprised your entire array--you might not be able to resilver the last failed disk before the next one went). The theory is that getting drives from different "lots" will help avoid this potential problem. Carried further, this theory would suggest that you get drives from different manufacturers. Rather than getting 6x WD Reds, maybe I should get two of those, two HGST NAS drives, and two Seagate NAS drives.

As I said, though there is some logic there, I don't (at least not intentionally) do this. I simply don't believe the chances of there actually being a problem, to which staggering your purchases would be a solution, are high enough to bother with it.

 

[Fuganater]

I am way too OCD to have different drives in my chassis.

Sent from my SM-T700 using Tapatalk

 

[solarisguy]

@Dat Sysadmin, your real problem™ is your decision to go with the WD Green model, and not using WD Red drives.

To buy from the same batch or not to buy from the same batch, that is a question that has a longer answer than yes or no; in short: it depends :D

If one is buying only 2-3 drives, then I would recommend buying each drive made in different weeks. When buying in a store that could be trivial or not (depends on a store). Mail-order is less than trivial, so I would buy from the same mail-order place spreading purchases by two weeks. Unless the 3 drives are to be in RAID-1, then I think it does not matter ;-)

If buying drives to be in RAID-Z2 or RAID-Z3, I would not be spreading purchases, unless I am buying 8 or more drives. A bad batch can happen to any manufacturer...

I would try to acquire hard drives that are the same model with the same firmware version. Then it is much easier to troubleshoot. And the pool performance is predictable. Let's say my golden standard are WD Red drives in RAID-Z2 that has 6 drives. If I had 3 WD Reds and 3 lousy drives, I would have a lousy RAID-Z2. If I had 3 WD Reds and 3 excellent drives worth their weight in gold, performance of my RAID-Z2 would be only that of WD Reds.

 

[solarisguy]

P.S. Manufacturing date is on the hard drive label. Take any drive into one hand and a magnifying glass into the other and take a close look.

 

[danb35]

@Dat Sysadmin, your real problem™ is your decision to go with the WD Green model,

Why is that such a problem? Run WDIDLE3.EXE on them and you're fine. You don't get the longer warranty of Reds, but that may or may not be a concern for him.

 

[solarisguy]

And the benefit of using a drive model that is discontinued and was never destined to be in a NAS is ?

 

[Deleted47050]

@Dat Sysadmin, your real problem™ is your decision to go with the WD Green model, and not using WD Red drives.

I think it really depends over what expectations you have over your system. I have always used WD Red drives in my servers and NAS boxes, but this time, for this type of media-serving machine that is likely going to be shutdown at night, I really don't see the point in investing on Red drives.

Plus, lots of people here in the forum seem to be using Green drives without problems, for probably much heavier workloads, so I am definitely curious to see how they work first hand.

Why is that such a problem? Run WDIDLE3.EXE on them and you're fine. You don't get the longer warranty of Reds, but that may or may not be a concern for him.

This. It is my understanding that this is the only real thing in which the two differ (besides warranty, as you mentioned). I am still trying to understand whether having purchased all of my Red drives has been worth it or not, especially after seeing that a big batch of Red drives came with the parking settings of the Green drives some time ago :/

 

[danb35]

And the benefit of using a drive model that is discontinued and was never destined to be in a NAS is ?

Price, to name an obvious benefit. What's the drawback? Discontinued? BFD. Computer hardware cycles faster than most, so anything you buy in this field will be discontinued before long anyway. The warranty's still good. Not designed for NAS? Again, IMO, BFD. I'm open to being shown that there's an actual, significant difference between "NAS" drives and "desktop" drives, but I haven't seen it yet, and I'm skeptical. Why spend 30% more (as of a minute ago at amazon.com, WD Green 1 TB are $50US; Red are $65) for no quantifiable difference?

 

[Bidule0hm]

Longer warranty is a good argument I think :)

 

[solarisguy]

Since the original question was about getting the best insurance, to me the benefit of a 3-year warranty over a 2-year one was self-evident.

P.S. When comparing WD10EZRX to WD10EFRX:
Load/unload cycles: 300,000 (Green) vs. 600,000 (Red)
MTBF (hours): 500,000-600,000 (Green, estimated) vs. 1,000,000 (Red)
Acoustics (dBA) Seek (average) for 1TB 3.5" model: 24 (Green) vs. 22 (Red) <<== I was expecting the opposite, and actually wanted to point to the noise data as an argument for the Green model...

 

[joeschmuck]

Since the original question was about getting the best insurance, to me the benefit of a 3-year warranty over a 2-year one was self-evident.

Me too but I like all the other stuff you added in but if I get 1 million hours before a failure, I'll be one happy camper :D. (Seriously, I do know how MTBF is calculated, we do that with parts we sell to the government and it always sounds like snake oil to me.)

 

[danb35]

I agree with the warranty benefit, though since you'll get a refurb drive back if you have to RMA it, I've seen a lot of folks hesitant about putting those into their servers. If you aren't going to trust the replacement drive, the warranty (whatever the length) loses quite a bit of value. As to the other specs...

My reds have been in my server for just under a year. They have a load cycle count of less than 20. At this rate, I have 15,000 years to go before I hit the limit for the Greens. Yes, the spec for the red is double this, but so what? The green MTBF says that, on average, they'll last 57 years of 24x7x365 operation. Again, the red spec is twice as high, but (1) this number is pure fantasy, and (2) if the number were meaningful, I'd be replacing drives long before then anyway. But as it is, I don't consider the difference between $NUMBER_PULLED_OUT_OF_ASS and 2 x $NUMBER_PULLED_OUT_OF_ASS to be really significant.

The acoustics are an objective, measured value, unlike the other two, but I don't know how significant the difference between 22 dB and 24 dB really is. But then, my server is in a 2U rack chassis, so fan noise overwhelms drive noise anyway.

 

[Bidule0hm]

but I don't know how significant the difference between 22 dB and 24 dB really is.

Not much. The human ear can differentiate levels differences of 1 to 3 dB (it varies from one person to another, some can differentiate only 0.5 dB but it's extremely rare) and each 10 dB the sound seems twice as loud.

 

[joeschmuck]

The green MTBF says that, on average, they'll last 57 years of 24x7x365 operation.

Eh? Nope, not really. MTBF is not a statement of how long a single drive will last. It's a statement on how long a large group will last before a single drive failure occurs. For instance WD takes 500 hard drives and powers them on. They run them and if there is no failure of any drive for 2000 hours, then they have a MTBF of 1,000,000 hours. Now if you want to shorten the testing time you could just take 1000 hard drives and test them for 1000 hours, same results. It's snake oil but it isn't 57 years otherwise someone would produce a warranty longer than 7 years (the longest I've ever heard of for a HDD).