Question about running multiple pre-clear passes

[Queball]

Has anyone identified problems with a hard drive by running multiple passes of the pre-clear script that were not found during the first pass?

 

Just curious about the importance/value of running multiple passes.

 

Thanks in advance for any replies.

[prostuff1]

Has anyone identified problems with a hard drive by running multiple passes of the pre-clear script that were not found during the first pass?

 

Just curious about the importance/value of running multiple passes.

 

Thanks in advance for any replies.

Yes

[aiden]

Has anyone identified problems with a hard drive by running multiple passes of the pre-clear script that were not found during the first pass?

 

Just curious about the importance/value of running multiple passes.

 

Thanks in advance for any replies.

Yes.  I always run 3 passes, even on 3TB drives.  I have found failures in the 3rd pass that didn't happen in the first 2.

[SCSI]

Yes.  I always run 3 passes, even on 3TB drives.  I have found failures in the 3rd pass that didn't happen in the first 2.

 

How long did the 3 passes take on the 3TB drive? I'm guessing 4-6 days?

[Johnm]

Yes.  I always run 3 passes, even on 3TB drives.  I have found failures in the 3rd pass that didn't happen in the first 2.

 

How long did the 3 passes take on the 3TB drive? I'm guessing 4-6 days?

36-40 hours each pass. depends on the hardware.

[bman]

Has anyone identified problems with a hard drive by running multiple passes of the pre-clear script that were not found during the first pass?

 

Absolutely, positively, and multiple times, yes

 

I use the preclear script to burn-in all the hard drives I want to use on servers both at home and at work.  Last time I had to send a LaCie external 2TB drive back because it came to a halt (logically not literally) sometime during preclear pass number 6.

 

I always pass each drive 7 times now, and it takes forever.  I do four at a time, but it's about 10 days' duration on a 2TB LaCie external over the eSATA port using a Dell Optiplex 745.

 

One pass isn't even enough to get the drive to run out of good sectors, if there are too many bad ones, and excess heat during the burn-in period will exacerbate any existing flaws.

 

My logic is simple - do it up front while the drive can still be RMA'd and you're not depending upon that drive to keep your data safe.

 

[Queball]

Thanks for all the replies.

 

One last thing, what's the modifier to run multiple passes?

[graywolf]

Thanks for all the replies.

 

One last thing, what's the modifier to run multiple passes?

 

-c 7

 

to run 7 passes

 

[Rajahal]

Thanks for all the replies.

 

One last thing, what's the modifier to run multiple passes?

 

-c 7

 

to run 7 passes

 

 

I hope that's a joke, or just an extreme example.  Running 2 - 3 passes should be sufficient in nearly all cases.  7 passes is way too many in my opinion.

[Johnm]

 

I always pass each drive 7 times now, and it takes forever.

 

yikes.. that is some serious wear and tear on a drive.

if it was not dead before, it might be after.

 

[johnm160]

Write speeds to my array grind to a halt during preclear so I could not imagine doing 3 passes per drive at this point. I have not been able to pin point the bottleneck either. I have been thinking about a second unraid box on an old pc just to do preclears.

[Rajahal]

Write speeds to my array grind to a halt during preclear so I could not imagine doing 3 passes per drive at this point. I have not been able to pin point the bottleneck either. I have been thinking about a second unraid box on an old pc just to do preclears.

 

The hardware in your sig looks plenty fast enough to handle a few preclears without slowing down the array.  Make sure your BIOS settings are correct (drives should be running in AHCI mode not IDE or combined mode).

[johnm160]

Write speeds to my array grind to a halt during preclear so I could not imagine doing 3 passes per drive at this point. I have not been able to pin point the bottleneck either. I have been thinking about a second unraid box on an old pc just to do preclears.

 

The hardware in your sig looks plenty fast enough to handle a few preclears without slowing down the array.  Make sure your BIOS settings are correct (drives should be running in AHCI mode not IDE or combined mode).

 

It is set to AHCI, transfer to the array during preclear was 3MB/s. Preclear was on the Masscool 2 port adapter and precleared at about 100/MB/s.

 

I am not sure if that cheapo card is hogging resources somehow.

[UhClem]

 

It is set to AHCI, transfer to the array during preclear was 3MB/s. Preclear was on the Masscool 2 port adapter and precleared at about 100/MB/s.

 

I am not sure if that cheapo card is hogging resources somehow.

 

2 questions:

 

1. Is there an array drive on the other port on the Masscool (the one not handling the preclear)? You clarified this in the "Re: preclear" thread.

 

2. How is the read speed of your array during this? Both, in general, and, if answer to #1 is yes, the read speed of that drive.

 

--UhClem

 

[johnm160]

 

It is set to AHCI, transfer to the array during preclear was 3MB/s. Preclear was on the Masscool 2 port adapter and precleared at about 100/MB/s.

 

I am not sure if that cheapo card is hogging resources somehow.

 

2 questions:

 

1. Is there an array drive on the other port on the Masscool (the one not handling the preclear)? You clarified this in the "Re: preclear" thread.

 

2. How is the read speed of your array during this? Both, in general, and, if answer to #1 is yes, the read speed of that drive.

 

--UhClem

 

 

I actually did not read from the array during that preclear, but I am going to preclear another drive tomorrow so I will test it and report back.

[UhClem]

 

I actually did not read from the array during that preclear, but I am going to preclear another drive tomorrow so I will test it and report back.

 

Good ... Another thing (I hope you see this) ... While your preclear is running, also do a (test) write to the array and have a "tail -f syslog" running. Look for any anomalies/errors. That write speed (~3 MB/s) is so atrocious that I suspect there will be some glaring clues from the kernel.

 

--UhClem

 

[johnm160]

 

I actually did not read from the array during that preclear, but I am going to preclear another drive tomorrow so I will test it and report back.

 

Good ... Another thing (I hope you see this) ... While your preclear is running, also do a (test) write to the array and have a "tail -f syslog" running. Look for any anomalies/errors. That write speed (~3 MB/s) is so atrocious that I suspect there will be some glaring clues from the kernel.

 

--UhClem

 

 

Thanks will give that a shot too

[johnm160]

Well preclear is in pre-read and I am copying to the array at 44MB/s I will see what happens once it starts to write zeros to the drive, but for now everything looks normal.

[johnm160]

Preclear is writing zero's to the drive at 97 MB/s and I am copying to the array at about 32 MB/s. I am not sure what went wrong the first time but clearly its much better now.

 

Edit to add:

 

I am actually copying to the array at around 37 MB/s apparently the mover was still running when I tested 32 above. I am reading from the array at 55 MB/s, no clue what I read at while preclear is not running. I don't think I ever checked that.

 

 

[UhClem]

... I am not sure what went wrong the first time but clearly its much better now. ...

Murphy was just jerking you around . ...He'll go f* with someone else now.

 

--UhClem

[johnm160]

... I am not sure what went wrong the first time but clearly its much better now. ...

Murphy was just jerking you around . ...He'll go f* with someone else now.

 

--UhClem

 

I have had my fill of murphy lately, he better not poke his head around anymore or he is libel to get hurt. 

[bman]

 

I always pass each drive 7 times now, and it takes forever.

 

yikes.. that is some serious wear and tear on a drive.

if it was not dead before, it might be after.

 

 

I pass each drive for 7 passes, too... and what would you know, the last Hitachi I did that to showed up perfectly until I tried to rebuild the array with it in place... then - wham-o.... 537642 sync errors, and SMART reports failed with hundreds of thousands of bad sectors.

 

I'm not sure if 7 passes is too many or not enough, but it's a real solid workout for a drive, and if it doesn't fail during 7 passes, it probably won't fail for a good long time (except for that last one that proved my theory wrong) 

[WeeboTech]

 

I always pass each drive 7 times now, and it takes forever.

 

yikes.. that is some serious wear and tear on a drive.

if it was not dead before, it might be after.

 

 

I pass each drive for 7 passes, too... and what would you know, the last Hitachi I did that to showed up perfectly until I tried to rebuild the array with it in place... then - wham-o.... 537642 sync errors, and SMART reports failed with hundreds of thousands of bad sectors.

 

I'm not sure if 7 passes is too many or not enough, but it's a real solid workout for a drive, and if it doesn't fail during 7 passes, it probably won't fail for a good long time (except for that last one that proved my theory wrong)   

 

preclear passes are not enough to exercise the drive.  Badblocks exercises the drive better then a dd write/read test of only 0's.

 

badblocks does 4 passes each with a specific bit pattern to try and catch marginal sectors.

 

In addition, sometimes these marginal errors do not show up unless there is other activity on the bus.

For example, if there is a marginal sector, the drive may pause when it reaches that sector.

If there are other operations present on the bus, then it may reveal the issue.

 

I mention this because of recent tests in my system where a badblocks test on 5 drives simultaneously revealed a questionable drive that did not reveal smart errors.

 

What was noticed is the drive was resetting itself on the bus. At that point I decided to take the drive out of operation.

 

[Joe L.]

 

I always pass each drive 7 times now, and it takes forever.

 

yikes.. that is some serious wear and tear on a drive.

if it was not dead before, it might be after.

 

 

I pass each drive for 7 passes, too... and what would you know, the last Hitachi I did that to showed up perfectly until I tried to rebuild the array with it in place... then - wham-o.... 537642 sync errors, and SMART reports failed with hundreds of thousands of bad sectors.

 

I'm not sure if 7 passes is too many or not enough, but it's a real solid workout for a drive, and if it doesn't fail during 7 passes, it probably won't fail for a good long time (except for that last one that proved my theory wrong)   

 

preclear passes are not enough to exercise the drive.  Badblocks exercises the drive better then a dd write/read test of only 0's.

 

badblocks does 4 passes each with a specific bit pattern to try and catch marginal sectors.

 

In addition, sometimes these marginal errors do not show up unless there is other activity on the bus.

For example, if there is a marginal sector, the drive may pause when it reaches that sector.

If there are other operations present on the bus, then it may reveal the issue.

 

I mention this because of recent tests in my system where a badblocks test on 5 drives simultaneously revealed a questionable drive that did not reveal smart errors.

 

What was noticed is the drive was resetting itself on the bus. At that point I decided to take the drive out of operation.

 

Even badblocks may have its flaws when used on a modern disk its internal RAM cache.  You say badblocks writes 4 patterns... I'm sure it does, but...

How many actually get written, and re-read from the physical disk platters, and how many are immediately satisfied from the disks internal 64meg cache memory. 

 

Unless the blocks written are large enough to ensure that the physical disks are always being written and read, the test is only getting as far as the memory in the disk drive.

[abs0lut.zer0]

sorry to jump in at the end of this conversation but does that mean that your preclear script could suffer from the same fate Joe L.

or does it force write and reads to every physical sector ?