Ancient Hard Drives

[bollar]

I have a server where old hard drives go to die. It's basically a failover machine that's meant to be used in the event that the primary server *and* the first replication fail. Just wanted to see if anyone can challenge the age on these drives!

Code:

+------+----------+----+-----+-----+-----+-------+-------+--------+------+------+------+-------+----+
|Device|Serial    |Temp|Power|Start|Spin |ReAlloc|Current|Offline |UDMA  |Seek  |High  |Command|Last|
|      |          |    |On   |Stop |Retry|Sectors|Pending|Uncorrec|CRC   |Errors|Fly   |Timeout|Test|
|      |          |    |Hours|Count|Count|       |Sectors|Sectors |Errors|      |Writes|Count  |Age |
+------+----------+----+-----+-----+-----+-------+-------+--------+------+------+------+-------+----+
|ada0  |CVCV238603|    |91705|     |     |      0|       |        |      |   N/A|   N/A|    N/A|3821|
|ada1  |5YD       | 32 |17248|   69|    0|      0|      0|       0|     0|     0|     0|      0|   0|
|ada2 !|9XW       | 34 |31448| 1043|    0|    126|     21|      21|     0|   447|     4|     33|   0|
|ada3  |6XW       | 33 |32619| 6962|    0|      0|      0|       0|     0|  6136|    64|4295163|   0|
|ada4  |6XW       | 33 |49130| 7018|    0|      0|      0|       0|     0|  3341|    25|      1|   0|
|ada5 !|6XW       | 32 |44704| 4299|    0|    391|      0|       0|     0|   471|   241|     55|   0|
|da0   |MK0331YHG | 40 |37360| 8806|    0|      0|      0|       0|     0|   N/A|   N/A|    N/A|   0|
|da1   |MK0311YHG | 40 |37439| 5531|    0|      0|      0|       0|     0|   N/A|   N/A|    N/A|   0|
|da2 ! |MK0311YHG | 39 |33466| 5578|    0|     37|      6|       0|     5|   N/A|   N/A|    N/A|   0|
|da3   |MK0311YHG | 38 |28373| 5471|    0|      0|      0|       0|     1|   N/A|   N/A|    N/A|   0|
|da4   |MK0311YHG | 39 |37637| 8882|    0|      0|      0|       0|     0|   N/A|   N/A|    N/A|   0|
|da5   |MK0331YHG | 40 |37340| 8772|    0|      0|      0|       0|     0|   N/A|   N/A|    N/A|   0|
|da6   |MK0331YHG | 40 |34896| 5864|    0|      0|      0|       0|     0|   N/A|   N/A|    N/A|   0|
|da7   |MK0331YHG | 38 |37355| 8756|    0|      0|      0|       0|     0|   N/A|   N/A|    N/A|   0|
+------+---------------+----+-----+-----+-----+-------+-------+--------+------+------+------+-------+----+

For those of you interested, this system is configured as two-six disk Z1 VDEVs plus a hot spare.

Ignore all of the data for ada0 -- it's an Intel SSD that reports power on hours in a different format.

 

[joeschmuck]

Age of the drive is ether the date it was manufactured or how long it's been powered on. Your oldest drive has been powered on for over 5.6 years. If I knew the model of the drive I might be able to look up the warranty for it and then figure out when it was manufactured but I'd need the full serial number for that as well but a guess would be in 2009 but that isn't very precise.

So where was the challenge?

EDIT: That old drive is still a SATA drive, not ancient at all, if you had an MFM or RLL connected drive, I'd be very impressed!

 

[Jailer]

Heh, I got an old Seagate 250GB drive that I use for all my alternate OS installs. It's been around the block a few times. ;)

And yes I know my drives are too hot. It's a bit warm here today and the hard drive cooling sucks donkey balls in the lian li v1000.

 

[Arwen]

Talk about age, I had a production server, (not FreeNAS/BSD), with a pair of
un-mirrored Seagate 72GB SCSI drives. You say, so what? Server was up 8.5
years!

Drive one had failed before I started with the company. All the application data
was moved to the OS drive, zero. Then drive zero started getting bad blocks. So I
replaced drive one. Then had to manually copy, (not mirror), the OS over to drive
one and then boot off drive one. I lucked out in that not one bad block was in use
by the OS.

Last, replaced drive zero and mirrored the disks. Good for another 8.5 years at least!
Except now I can hot swap the disks...

 

[bollar]

The drives on this system are either Seagate Barracuda LP 2TB or Hitachi Ultrastar 3TB. And of course, they're all well out of warranty.

I aspire to get some of the power-on hours you've reported!

 

[diedrichg]

I'm now in the 1% club!

I have (4) 2TB WD Reds in RAIDZ2 - yes, I now know I would have twice the storage with just two more drives, hindsight is a bi+ch though. The difference in power-on hours is because I replaced (4) 1TB Blue drives one-by-one as I had the cash. We strictly use our FreeNAS as a file server for our personal files, photos, software and music so we currently don't require much room.

Two drives at 1% of their lifespan...
ada0 = 10,547h / 24pcc (power cycle count)
ada1 = 6,042h / 13pcc
ada2 = 6,040h / 13pcc
ada3 = 11,625h / 32pcc

All drives run at a cozy 27°C (80°F) in our cool basement. Yes, all drives have been set to /D using wdidle3.

 

[joeschmuck]

My drives are WD Red 2TB drives. Of course I started out with 4 drives and added 2 more later as you can see by the hours running. My first set of drives warranty expires Oct 2015, the last two expire March 2016. They are going strong and I hope they last a few more years. I do power off my unit about three times a year to blow the dust out of it. It's in the basement too, staying cool.

ada0= 24,983h
ada1 = 24,964h
ada2 = 24,979h
ada3 = 21,076h
ada4 = 21,530h
ada5 = 24,988h

Two drives at 1% of their lifespan...

Not sure I understand what you are saying, what drives are at 1% lifespan? The WD Red 2TB drive has a 3 year warranty which equals ~26,280 runtime hours maximum.

 

[diedrichg]

Not sure I understand what you are saying, what drives are at 1% lifespan? The WD Red 2TB drive has a 3 year warranty which equals ~26,280 runtime hours maximum.

MTBF 1,000,000. Of course you have now just jinxed me.

 

[joeschmuck]

MTBF 1,000,000. Of course you have now just jinxed me.

LOL, yea, 114.15 Years! ROTFL.

And you know how they come up with those specs right? They take lets say 500 hard drives, run them for 2000 hours and if there are no failures, it must mean they can run for at least 1,000,000 hours before a failure occurs, which translates into 114.15 Years! It's a stupid way to rate the MTBF. In my company we actually use real data to rate MTBF based on component historical failure rates and on a new product, we just make a best guess based on previous similar components. I'd say most of the time our prediction is on the safe side until actual failures come in, and it does happen. Like a hard drive by far isn't 1,000,000 hours, we use something closer to 5 years because it based on actual failure rates.

But in reality the warranty period is a better realization of how long a drive should be expected to last. If it lasts longer then you are exceeding what the manufacturer really believes they will last, at least from a financial sense.

 

[diedrichg]

Clearly I didn't think that one through. Point taken.

 

[ChriZ]

@joeschmuck are you implying that drives usually fail shortly after their warranty expires?
Nooooooo.....:p

 

[joeschmuck]

@joeschmuck are you implying that drives usually fail shortly after their warranty expires?
Nooooooo.....:p

Well I hope not, although I'm prepared to purchase 4 more drives next month, I'd rather wait another year. I'm drooling over a new computer system to build and I'll need every drop of money I can get my hands on. A nice small ITX form factor, Skylane CPU, 32GB RAM, 1TB SSD connected via PCI-e (that's the 2100MB/sec transfer speeds baby!), and silent. I will likely need an AIO liquid cooler for the CPU and hopefully two fans maximum. Small, Quiet, Lethal ! Oh, and a new keyboard. I can barely make out the keys on my keyboard. I want one with the MX Brown keys but the key caps must be laser etched so the letters never rub off again.

Well I'm dreaming of course and planning and scheming.

 

[cyberjock]

All of my old 2TB WD Greens have over 35k hours.. of the lot that I bought, only 2 failed before I retired the disks. :P

 

[Spearfoot]

I've got an old WD Caviar 26400 EIDE drive that was manufactured on August 26th, 1998. It still worked the last time I checked it back in May of '14. Yes, it has a whopping 6448.6 MB capacity!

 

[cyberjock]

Eh, I still have my "big" 1.2GB drive I had in high school (circa 1996). Still works too!

 

[Spearfoot]

Eh, I still have my "big" 1.2GB drive I had in high school (circa 1996). Still works too!

Ha, I've probably got the same model! But I don't believe mine works any more...

 

[cyberjock]

Yep.. my sticker looks just like that one. :P

 

[joeschmuck]

Eh, I still have my "big" 1.2GB drive I had in high school (circa 1996). Still works too!

High school 1996, damn I feel old, I was punching holes in the ocean on my last boat. My first hard drive was a 5MB ST-506 and I bought two of these used. They worked great. Keep in mind back in 1984 those were still a lot of storage. My computer was I believe an 80286, heck it could have been an NEC V20 running at a whopping 12MHz (a personal modification I made where you actually needed to replace soldered components to make frequency changes) and that was a huge leap over the 4.77 MHz of the IBM XT of the day (8080 CPU) I went though a lot of upgrades in a very short period of time. Nope, I don't still have those drives either. Good times.

 

[cyberjock]

My first hard drive was a Seagate 20MB RLL drive. :P

 

[Spearfoot]

@joeschmuck, you and I must be of an age... The first HD I ever bought was the 10MB version of that same ST-506 drive, for an original IBM-PC equipped with a whopping 512k of RAM and not one, but two full-height 5 1/4" floppy drives. We added an AST Six Pack Plus to max out the DOS-addressable memory to 640k and have a 384k RAM drive! Nirvana!

Armed with Turbo Pascal and using that primitive equipment, I wrote programs to drive plotters and digitizing tablets, complete with interrupt service routines to handle the RS-232 communications, which could easily outrun the CPU without proper buffering. We would upload FORTRAN programs to Boeing Computer Services in Seattle and run them on their supercomputers -- took longer to upload the code and data and download the results than it did to actually run the programs!

We really have come a very long way indeed, these last thirty years.