Reallocated Sector Count Warning!?! CAVIAR GREEN DRIVES Problems

[src666]

For any always-on system, you need to consider green drives as throw-away drives. They will work for some period of time, but they are also going to die sooner rather than later. The refurb drive you (might) get under warranty will die even faster.

As long as you don't mind tossing drives, and whatever data is on them, every year or two, then they are fine.

For me, the hassle (down-time, my time) of doing that is a far higher price than paying for good drives up-front. And if you don't believe me, I can include a picture of the stack of about 20 dead drives I have sitting on the shelf.

In my house, I have at least 25 drives spinning nearly full time. I lose between 3-5 a year. Anything I can do to lower that failure rate is worth the price.

And trust me, the cost of dealing with the WAF failure that comes from my wife losing a season of her favorite shows is MUCH worse than the checkbook pain of buying good drives. Sadly, buying the good drives doesn't guarantee you won't get any lemons. It happens.

And yes, there are always people who have personal experiences that differ, who have the cheapest drive they could find still running after years of abuse, who have all the luck in the world and never lose a drive. If that's you, congratulations. Doesn't change the fact that green-level drives are meant for low-uptime, low-volume, low-demand situations.

[src666]

Originally Posted by dcardellini

RAID card mfgrs recommend sticking to the same mfg and model in the array. I have no idea what my exposure is pushing this boundary....as I do not understand the underlying reasons.

This is mainly for performance reasons. Different drive manufacturers have different drive geometries, different cache sizes and caching algorithms, different pre-fetch strategies, etc. When you mix and match drives, your raid performance will not be optimal, since the array runs best when all of the drives are working in sync - the heads are moving at the same time, to the same location on the platter, the caches are flushing at (about) the same time, the pre-fetch is the same, etc.

On the other hand, with the huge ECC issues presented by today's large drives, they won't be in sync anyway. One will have to repeat a read to try to get good data, one will be reading a different sector, because it went south and had to be relocated into the spare area, etc.

Same manufacturer/same model gives you the best chance at optimal performance, but unless you are running some really hairy high-throughput application, different drives won't matter quite so much.

[dcardellini]

Savage and SRC:

Thanks for responses.

Yes, do auto verify.

My 17-drive array is really 16 drives in RAID 6 with the 17th as a hot spare.

The drives have been bought from at least 4 different places/different times...but now that I think about it, the oldest ones are the ones that have died, and they may have been from the same batch.

Never considered that my SuperMicro hot-swap 5-in3's could be at fault. I guess the minute I get another failure that falls on the same exact slot (or maybe same 5-in-3), I should be suspicious of this. However, the two drives I have replaced into the same slots are running flawlessly now for a few weeks with no Reallocated Sector warning messages.

Thanks for verifying that mixing drive models may not be a killer....as I really am not concerned with optimal performance.

Next questions:

1. As WD seem to be disabling TLER changes on some drive models, are the new Caviar Blacks still capable of being TLER'ed? Also, the Samsung noted above.....how does it handle TLER?

2. I am confused as to the actual usage profile that pushes these Green drives to early failure. I keep hearing "not for use in RAID applications."

My question is, for the way I use these (very light duty), written to once a day for about 10 minutes (a few big files), and read from pretty seldom....

How is my usage pattern in my RAID any different than a single disk used in a media application? Is the RAID doing some terrrible thing? I watch the little green lights on my chassis.....I watch the kW used on my UPS I can see these things are idle for long periods ( the RAID is not hitting on them unnecessarily).

From where I am sitting......it is just the fact that these Green drives stay on 24/7, and has nothing to do with the fact they are part of a hardware RAID.

....and if that is the case, then anybody planning to use these for even light duty media serving will be in trouble if they leave their NAS on 24/7.

If I am misusing these drives, then I accept responsibility for pushing the limits. However, if just leaving them on, idle 24/7, constitutes abuse, then WD must be getting swamped with failures about right now, and they will have lost me as a customer, for sure.

[davephan]

I have 9 Western Digital 2 TB drives, and had some trouble with the drives, but I don't consider the drives to be a bad design to be avoided. The drives are Western Digital 2 TB green WD20EARS and WD20EADS, probably similar to your WD 2 TB green drives.

Perhaps the problem might be heat related with people that are having high drive failure rates. My WD 2 TB green drives are running between 25c to 30c, except one that is running at 44c. My 9 drives have been all been running now for 6 months to over a year without problems. Each drive is mounted in Thermaltake iCage A2309 drive cooling cages, except for the hot drive. The product used to be available from Newegg, but is no longer available. It is still available from Amazon. The density is much lower than the 5 in 3 cages, so the airflow should be better, and it probably keeps the drives cooler than the high density cages. I tried adding a fan near the hot drive, but it did not lower the temperature at all. Maybe the drive cage will help cool down that drive or maybe that drive has a problem and might need to be replaced.

Of the 9 drives, I had trouble with 3 drives. Two drives failed when connected to a bad power supply. It turned out the one of the connectors on the power supply was supplying bad voltage, I assume too high voltage. I did not measure the voltage. Both 2 TB green drives immediately sparked and had a burnt smell after that. Both drives were immediately dead. That was a bad power supply problem. I replaced both drives that were destroyed by the bad power supply and I also replaced the bad power supply.

I also had one drive that ran for a day and went bad. No data was lost since it was setup with RAID 1. I returned the drive to the store and the replacement has been working fine since. Of the 3 drives with problems, I only consider one drive to be a drive problem, since the bad power supply was to blame for the other two drive failures.

Here is a link to the Thermaltake iCage A2309 drive cooling cages:
http://www.amazon.com/gp/product/B00...ef=oss_product

I suspect that the drive problems are for two reasons. The green drives are marginal for array use, since they are not recommended by the manufacture for that purpose. I also think that people with a large number of drives may be packing them too close together, with the physical density too high, which does not allow enough air flow for cooling. If you are experiencing many drive failures, I think you might try reducing the physical drive density to improve the airflow and lower the drive temperatures.

The WD advanced format drives require a jumper between 7 & 8 before putting the drives into an unRAID array. I didn't know about the required jumper, so I am moving many 8.6 TB of data off my unRAID server so I can redo the unRAID server with drives that have pins 7 & 8 jumped. Then I will move that data back to the unRAID server.

I have a replacement warranty from the store where I purchased them. So, if the green drives fail, they will be replaced with new drives, not refurbished drives. If I did not have the warranty, I would be tempted to replace each failed green drive, as they fail, with an enterprise 2 TB drive at twice the cost, $250 per drive. However, after the initial drive failures, the WD 2 TB green drives have been reliable for 6 to 12 months (the drives were purchased gradually over time).

Dave

[Fuzzy]

I'd really recommend the EVDS drives. (the AV-GP series). They have the power saving advantages of the Green drives, but have higher build quality and reliability. And, really.. when you think about it, they are designed specifically FOR this type of environment.

[Savage1701]

Originally Posted by dcardellini

Savage and SRC:

Thanks for responses.

Yes, do auto verify.

My 17-drive array is really 16 drives in RAID 6 with the 17th as a hot spare.

The drives have been bought from at least 4 different places/different times...but now that I think about it, the oldest ones are the ones that have died, and they may have been from the same batch.

Never considered that my SuperMicro hot-swap 5-in3's could be at fault. I guess the minute I get another failure that falls on the same exact slot (or maybe same 5-in-3), I should be suspicious of this. However, the two drives I have replaced into the same slots are running flawlessly now for a few weeks with no Reallocated Sector warning messages.

Thanks for verifying that mixing drive models may not be a killer....as I really am not concerned with optimal performance.

Next questions:

1. As WD seem to be disabling TLER changes on some drive models, are the new Caviar Blacks still capable of being TLER'ed? Also, the Samsung noted above.....how does it handle TLER?

2. I am confused as to the actual usage profile that pushes these Green drives to early failure. I keep hearing "not for use in RAID applications."

My question is, for the way I use these (very light duty), written to once a day for about 10 minutes (a few big files), and read from pretty seldom....

How is my usage pattern in my RAID any different than a single disk used in a media application? Is the RAID doing some terrrible thing? I watch the little green lights on my chassis.....I watch the kW used on my UPS I can see these things are idle for long periods ( the RAID is not hitting on them unnecessarily).

From where I am sitting......it is just the fact that these Green drives stay on 24/7, and has nothing to do with the fact they are part of a hardware RAID.

....and if that is the case, then anybody planning to use these for even light duty media serving will be in trouble if they leave their NAS on 24/7.

If I am misusing these drives, then I accept responsibility for pushing the limits. However, if just leaving them on, idle 24/7, constitutes abuse, then WD must be getting swamped with failures about right now, and they will have lost me as a customer, for sure.

I've had poor luck with 3 into 5 cages, 4x2.5x5.25" cages, and even backplanes. I've had arrays fail, pulled the drive from the cage/backplane, and the drive/array immediately rebuilds itself. I think most are cheaply made. I like iStar and SuperMicro.

That TLER utility is floating around out there. Opinions vary as to whether or not one should mess with it. I don't know if it is universal to all WD drives.

As far as I know, RAID is going to verify a lot, which is going to give you a ton of load/unload cycles which is hard on Green drives that go for max saving by unloading the heads and powering down, which will drive any RAID controller nuts unless it is implementing its own spin-down "MAID" protocol (think Areca). And a lot of times if a drive is marked bad, the array won't take it back.

I've been using Spinpoint F4's with good luck so far.

I know from talking to techs at 3Ware that their controllers squeeze everything they can out of a drive, so in that sense, the controller is demanding of the drive. But the controller does not "know" the drive is not up for the challenge, so the RAID people get blamed for "beating up" on drives that maybe ought not have been there in the first place. The nightmare scenario is ECC errors on 3 disks under RAID 6, so the Hot Spare can't be implemented. I think ECC errors are more common in cheaper drives. Verify if you can.

I can tell you my SpinPoints read cooler under SMART by quite a bit.

I've used some 640 & 750 GB laptop drives with good success.

Also, consider SuperSpeed's SuperCache app. I can take a 50MB/s 2.5" drive and get sustained writes of up to 50% higher just for allocating some RAM to let pages sit dirty while the heads fly around.

Also, as I said before, 64K blocks and I keep my drives set to fill up from the edge to the center, with wee hours in the morning defragging by file name.

I'm seriously considering an iStar unit with 4 2.5" 1TB units set to older 512byte sectoring and seeing how my 3Ware system can handle it as a RAID 5 array. If I can defrag and use SuperSpeed on it and get sustained 130-150 R/W activity, my needs don't go much beyond that. I could cut my power use substantially, as well as my space being used. Love to test it. I think some of the 2.5" units are really coming into their own for this sort of thing. And they are easier to cool in many situations.

I guess if price were no object, get a bunch of 1TB SSD's. Write archive material to them and just read from them. They SHOULD last forever...Maybe in 3-5 years I'll try that.

[Savage1701]

Originally Posted by src666

For any always-on system, you need to consider green drives as throw-away drives. They will work for some period of time, but they are also going to die sooner rather than later. The refurb drive you (might) get under warranty will die even faster.

And trust me, the cost of dealing with the WAF failure that comes from my wife losing a season of her favorite shows is MUCH worse than the checkbook pain of buying good drives. Sadly, buying the good drives doesn't guarantee you won't get any lemons. It happens.

OK, I gotta know - what is "WAF" an acronym for? I can't figure it out, even though I've dealt with that same scenario with my wife.

[PhilH]

OK, I gotta know - what is "WAF" an acronym for? I can't figure it out, even though I've dealt with that same scenario with my wife.

WAF = Wife Acceptance Factor

And sometimes some people also say it stands for:

WAF = Wife Approval Factor

[Savage1701]

Oh, yeah, OK. That makes perfect sense now. :-)

Appreciate the clarification!

Thanks!