How many pre clear cycles is enough?

[ashman70]

I am currently doing a pre clear on a Seagate 8TB disk that will replace my current 5TB parity disk. I realize this is a question that comes down to personal preference however I am interested in what people have to say about it.

I am probably about half way through the pre clear cycle, should one cycle be enough or have people actually found issues on a second or third cycle after the first completed without errors? I realize there is nothing to lose but time by doing more cycles.

 

Thanks

[gundamguy]

IMO: I do one cycle, and that's fine with me.

 

I'm not as worried about it as some people.

[garycase]

As you've noted, it's a personal preference.    I test my drives using WD's Data Lifeguard, doing two complete quick & extended tests; with a full write zeroes in-between.    Then I do a preclear with the -n option if it's a drive I'm adding to UnRAID [skips the pre- and post- reads, so it's MUCH quicker than a full cycle].

 

I think one cycle with ZERO issues is fine -- if there are ANY issues I wouldn't bother with another cycle ... I'd simply return the drive.

 

[Frank1940]

I would personally favor putting about a hundred hours on the drive.  That should be enough to get it by the period when most infant morality occurs. 

[Chris Pollard]

I tend to do one only but then I have the drive as a cache in my test server so it gets some limited real world use before putting it in the array.

[John_M]

For me it's a short SMART self-test, one pre-clear cycle, then a long SMART self-test. If it fails the short self-test it's rejected immediately. If it fails the pre-clear it's given another chance (another cycle) before being rejected. If it fails the long self-test it's rejected. I take notice of the SMART reports early in the disks' lives and keep an eye on them throughout. Different people have different strategies - you'll find out what works best for you. I keep some spare disks pre-cleared and ready for use either as replacements or as new additions or in new builds. If I was to find myself in the situation where I needed to replace a failed disk but didn't have one pre-tested and waiting I would use a virgin rather than leave my array unprotected for longer than absolutely necessary.

[RobJ]

There are a very few reports that actually had a successful first Preclear pass then it failed on the second or not even until the third.  So it's possible, but rare.  Therefore, I only do one pass, and figure that probably finds ~99% of early drive failures.  Once it's in the unRAID array, the parity protection will cover any later failures.

[trurl]

I've always done only 1 pass.

[ijuarez]

I do 3

[ufopinball]

I do 3

 

I just did 3 cycles on some 4TB drives.  Each cycle ran 37ish hours, so that breaks the 100 hour testing mark, too.

[ijuarez]

I do 3

 

I just did 3 cycles on some 4TB drives.  Each cycle ran 37ish hours, so that breaks the 100 hour testing mark, too.

 

i did 2TB 4 at time 96+ hours

[ashman70]

I am just curious, for those doing more than one cycle, what is the thinking behind doing a second and a third? Is it hitting the 100 hour mark or just that three is better than one logic?

[Frank1940]

I am just curious, for those doing more than one cycle, what is the thinking behind doing a second and a third? Is it hitting the 100 hour mark or just that three is better than one logic?

 

My personal objective is just to get about a hundred hours on the drive and the preclear cycle does that while stressing and testing the drive at the same time.  I would rather have a drive fail in the testing cycle then while attempting to rebuilt data on it.  (As a point of disclosure, I have have never had a drive fail during a preclear operation or at any time that could be considered a infant morality failure.  All of my failures have been after a year or more of operation.)

[trurl]

My (somewhat post hoc) justification for only doing one preclear is because I typically do the preclear for testing, then use the disk to upsize by rebuilding a smaller disk, then do a non-correcting parity check. So preclear does a full read, a full write, a full read, then rebuild does a full write, then parity check does another full read.

[BobPhoenix]

I've had several DOAs on the second or third cycle but made it through the 1st just fine.  I've had very few drives die after the 3rd cycle after they are in the array.  At least none that have done so within a short time of being put into an array.  Most of my failures after 3 cycles have been after 1-3 years of use.  I would do 6 cycles if I though it would make a big difference but 3 cycles appears to catch 95% of the bad drives.

[John_M]

I think that three pre-clear cycles is excessive for consumer grade disks. OK, some datacentre disks are rated at 500 or so TB/year and a typical workload for NAS disks is 180 TB/year but look at the spec sheet for Seagate desktop disks and you'll see that the workload rate limit (as defined here) is a mere 55 TB/year. Three pre-clear cycles (read, write, read, write, read, write, read) shifts seven times the disk's capacity or, in the case of the latest and greatest ST8000DM0002, 56 TB. That means that it's annual workload is exceeded before it has even been put to "profitable" use. It is, quite literally, being pre-cleared to death.

[JonathanM]

I think that three pre-clear cycles is excessive for consumer grade disks. OK, some datacentre disks are rated at 500 or so TB/year and a typical workload for NAS disks is 180 TB/year but look at the spec sheet for Seagate desktop disks and you'll see that the workload rate limit (as defined here) is a mere 55 TB/year. Three pre-clear cycles (read, write, read, write, read, write, read) shifts seven times the disk's capacity or, in the case of the latest and greatest ST8000DM0002, 56 TB. That means that it's annual workload is exceeded before it has even been put to "profitable" use. It is, quite literally, being pre-cleared to death.

Which is why I don't buy Seagate drives. 

[John_M]

Seagate's 8 TB disks are the only ones that even come close to being affordable. It's also the only company I've found that publishes the annual workload of its consumer disks. I don't expect WD's figures to be much different, but then WD doesn't publish specs of any sort these days.

[JonathanM]

Seagate's 8 TB disks are the only ones that even come close to being affordable. It's also the only company I've found that publishes the annual workload of its consumer disks. I don't expect WD's figures to be much different, but then WD doesn't publish specs of any sort these days.

I quit buying WD as well. I like HGST's consistent statistical performance in the backblaze environment. I realize it's not a one for one comparison to an unraid environment, not even close, but the consistently good numbers have to mean something. As far as affordability, my time and data is worth the extra money. Yours may not be, but that's a personal decision.

[John_M]

As far as affordability, my time and data is worth the extra money. Yours may not be, but that's a personal decision.

 

I wish I could afford justify the cost of HGST's 8 TB Helioseals, but I can't so I'll make do with Seagate's air-filled drives.

 

Since Backblaze has been mentioned, you might be interested to read this update. If you believe such things, it seems that Seagate has much improved, WD has much worsened and HGST has retained its lead, with the statistically insignificant Toshiba taking second place.

 

Edit: I haven't been able to find the annual workload limit figures for HGST disks either. But in the case of WD enterprise class disks I believe the mysterious SMART parameter 16 to which I refer here is measuring that very thing.

 

[BRiT]

3 passes required

 

I have had drives pass cycle 1 and 2 but die during cycle 3. I have not had any drives which passed the 3 cycles die in under 4 years. If they survived the 3 cycles they lived out the rest of their typical lifetime.

[trevisthomas]

This is a great thread.  I'm especially intrigued with the comments about drives having an annual workload limit!  Fascinating.

 

I usually preclear only once but I am trying to have hot spares in the array now so that a disk failure can be replaced immediately.

 

I can't think of ever having disk die before it was a few years old and more often than not they drives are still fine but since their storage limits have been vastly surpassed that they're just wasting energy. 

[John_M]

This is a great thread.  I'm especially intrigued with the comments about drives having an annual workload limit!  Fascinating.

 

I've only been able to find figures for Seagate disks but as a number of people's sigs say that they are using STx000DM disks in their arrays I thought it only fair to point out the distinct possibility of pre-clearing them to death. It said in one of the Seagate documents that the disk firmware records the workload so it's possible they could be reluctant to issue an RMA in cases where the figure is grossly exceeded. I know from Seagate documents that their Archive and NAS disks are rated at 180 TB/year and their Enterprise disks at something like 550 TB/year. I'd very much like to know the figures for other brands and models but the information is simply not published. Even the rotational speed is a secret in some cases.

 

While I've had laptop hard disks fail in service (they have a much harder life than desktop disks) over the past ten years I've replaced far more 3.5-inch disks that were working just fine than ones that had actually failed. Two reasons, either they were system disks in servers or workstations that were replaced with SSDs or they were data disks that were deemed to be too small.

 

The hard disk industry seems to be holding its breath at the moment. Apart from the recent release of conventional (non-SMR) 8 TB disks in just about every category by Seagate, increases in capacity seem to have stalled. I'm surprised that WD doesn't have anything bigger than 6 TB at the moment and I'm surprised that Seagate hasn't pushed SMR technology a little further.

 

[John_M]

I have had drives pass cycle 1 and 2 but die during cycle 3.

 

Seagate Desktop (DM) disks, by any chance?

 

[Frank1940]

This is a great thread.  I'm especially intrigued with the comments about drives having an annual workload limit!  Fascinating.

 

I've only been able to find figures for Seagate disks but as a number of people's sigs say that they are using STx000DM disks in their arrays I thought it only fair to point out the distinct possibility of pre-clearing them to death. It said in one of the Seagate documents that the disk firmware records the workload so it's possible they could be reluctant to issue an RMA in cases where the figure is grossly exceeded. I know from Seagate documents that their Archive and NAS disks are rated at 180 TB/year and their Enterprise disks at something like 550 TB/year. I'd very much like to know the figures for other brands and models but the information is simply not published. Even the rotational speed is a secret in some cases.

 

While I've had laptop hard disks fail in service (they have a much harder life than desktop disks) over the past ten years I've replaced far more 3.5-inch disks that were working just fine than ones that had actually failed. Two reasons, either they were system disks in servers or workstations that were replaced with SSDs or they were data disks that were deemed to be too small.

 

The hard disk industry seems to be holding its breath at the moment. Apart from the recent release of conventional (non-SMR) 8 TB disks in just about every category by Seagate, increases in capacity seem to have stalled. I'm surprised that WD doesn't have anything bigger than 6 TB at the moment and I'm surprised that Seagate hasn't pushed SMR technology a little further.

 

I am NEVER had any Hard Disk manufacturer refuse to issue an RMA and the replacement hard drives have always been shipped on the same (or next) day after receipt.  So, basically, they are not even looking at them until much later.  However, I also have no doubt that they do maintain records of returns and the names of those who are returning them.  If they find reason to suspect that a individual customer is abusing the RMA system, they may take some action to assure that they are not replacing drives that have not actually failed but rather are just at the end of their warranty period. 

 

If anyone has had an experience to the contrary, I would appreciate if that those persons post about the problem and the manufacturer involved