Is the system available immediately after replacing a smaller drive with a larger one?  I upgraded from 4.6 to 5rc 12 in order to use 3TB drives.  I have the parity installed and now i want to replace a 1TB drive on the array with the old 2TB parity drive.  I have pre-cleared the drive but i'm wondering if the system will be available during the reconstruction if i swap the 1TB data drive for the 2TB one?

 

Sorry if this has been asked and answered, i've had no luck finding it.

 

Thanks.

Yes, the array is available during that time => but performance will be notably degraded; and the rebuild will also take longer.    It's best to do the rebuild at a time when you simply don't need the array 

 

By the way, there was no reason to pre-clear the 2TB drive, unless you just wanted to test it thoroughly (although since it was taken from the system, it had clearly already been well-tested).

 

I wasn't sure if a new drive needed to be cleared or not so i went ahead and did it.  Good to know though. So if a drive is well worn would you not bother with preclearing unless it's parity? Or is it just completely not needed if the drive is broken it/known to be good?

If the drive is well "broken in" ... i.e. "known to be good" ... there's no reason to pre-clear it.

The pre-clear process prepares a drive to be ADDED to the array.    When you subsequently (after the pre-clear) add it, it's instantly added, and only needs to be formatted with the file system (a few minutes process).

 

But if you're using it as the parity drive, or to REPLACE a drive, then it doesn't save any time at all.   

Anytime I take a used drive and I'm going to move/use it in unRAID I do a preclear or a badblocks test.

 

If you have ANY pending sectors, the drive could cause you lots of grief in the future if you have to rebuild another drive.

 

if during the rebuild of any other drives, a sector off one of the other drives is unreadable and the drive goes off line, you could possibly loose the missing drive.

 

Preclear insures that every sector is written to. If any sector was in pending sector status, chances are it would be reallocated during the write and the substitute sector would be used from then on.

 

As unRAID'ers  We have the unique opportunity to test each drive and repair it should any sectors start going bad.

 

FWIW, I wouldn't use the array, except for read-only operations, while a drive was being rebuilt.

Thanks for the info.  I'll wait until tonight to do the switch.

FWIW, I wouldn't use the array, except for read-only operations, while a drive was being rebuilt.

 

Agree -- as I implied above 

In fact, I don't use the array for ANY purpose during a rebuild -- or, for that matter, during a parity check.  Yes, you CAN use it during these operations -- but the subsequent disk thrashing that any other activity causes simply slows everything down, so it's best to do those operations during times you can do without the array (e.g. overnight).

 

What's missing about the "broken in" drive comments is validation that all sectors have been utilized and there are no pending sectors.

 

if you are going to do a SMART long test. Then check the SMART stats for any pending sectors, you can just add it to the array.

 

Here's the thing to consider.

 

if the drive has some other operating system data on it.

The drive must be cleared to be "just added" to the array.

if it is replacing a drive, I believe you can just put it in place and write/upgrade to it.

 

if adding a new drive, and you do not "pre" clear it. unRAID will preclear it (unless it is the parity drive).

while the unRAID gui emhttp is clearing the drive, the array will be offline.

 

So it's best to do a preclear if there is other operating system data on the drive.

It's wise do to either preclear a drive before putting it in place or doing a SMART long test and validating the smart data.

 

For me, I do a badblocks 4 pass read test. Then a SMART long test (if the drive was already precleared).

 

If not, then I do a 4 pass badblocks write test, then a smart long test and compare the SMART data.

After that I'll do a preclear even though the last write test was all 0's anyway.

 

After you loose the ability to rebuild a failed drive due to a pending sector, you become a lil paranoid.

 

I also tend to be a bit paranoid about testing -- when I get a new drive I run WD Data Lifeguard; do a quick test; then an extended test; then a full write zeroes;  then repeat the first two tests; then run Seagate's SeaTools and run the SmartTest, Short Drive Self-test, Short Generic, and Long Generic tests.    THEN I attach the drive to one of my UnRAID boxes and run 3 passes of Pre-Clear !!    The process takes ~ 1 week/drive.

 

If there are ANY errors at any point in the process;  or ANY reallocated sectors on the drive; I return it 

 

In this particular case, however, I didn't think the drive needed additional testing, since it had been the parity drive ... thus was exercised a LOT during writes; and comprehensively during every parity check.    However, I should have suggested checking the SMART status 

 

Yes as a parity drive, the drive was "well worked".

I should have taken notice to that better.

 

My comments were more generic pertaining to any drive I add as data drive to the array.

 

Gary, that's one long battery of tests. I don't go that far.

 

For a new drive

I'll usually capture the drive's SMART status.

Then do a long test. and capture that.

I love seeing the first SMART long test in low hours.

After that I do The 4 pass badblocks read/write, smart long then preclear. Capture the logs and review.

It's close to a week on the larger drives.

 

At least we know our drives are good!

I think that something is being missed here ... yes, any drive, particularly parity, which has been in the array for a long time can b considered to be well proven, but ....

there will be sectors on the parity drive which have only been accessed during parity checks and parity builds, and may only have ever been written once in the life of a (data) drive.  Preclear guarantees to write and read every single sector several times.

 

One preclear pass is still a good idea for any drive about to be re-added to the array.

At least we know our drives are good!

 

Yes, I feel fairly confident about that 

I used to also run a Level 2 Spinrite test ... but decided enough was enough !!

 

I HAVE had a fairly high "fail" rate on WD Greens -- nearly 1 in 5 doesn't make it through that battery of tests -- BUT those that do have been exceptionally reliable ... in fact I've NEVER had one fail (out of ~ 100 or so on my systems and those I've built for friends).

 

The WD Reds have fared much better -- out of about 15 I've bought, NONE have had any issues (although there are a fair number of folks who haven't been so lucky on this and other forums).

 

By far my current drive of choice is the WD Reds.  [Although if they don't come out with a 4TB unit fairly soon I'll have to use the Seagates  ]

 

PeterB => I don't disagree that a Pre-Clear isn't a bad idea;  but the parity drive, assuming a reasonable number of parity checks have been done, has been very thoroughly worked out.    If it doesn't show any re-allocated sectors after extensive use in that position, I don't think additional testing is necessary (although it certainly doesn't hurt).    If you REALLY want to exercise a drive, do a Level 3 or 4 Spinrite run => that does far more testing than pre-clear (but also takes a VERY long time on a modern 3-4TB unit [over a week for Level 5].