Started my first pre-clear at 3:15 PM today.  The pre-read ended at 9:30 PM. Elapsed time 5:15.  Disk is a Hitachi 2 TB 7200 RPM unit. When it started it, reported a transfer rate 0f 119 mb/s. When it was near the end (98% complete on the pre read) it reported 64 mb/s.  I suspect this has to do with "cylinder" diameter and linear velocity of a rotating medium. It is now once again running at 119 mb/s at 2% complete on the writing of zeros.

 

All errors that were reported in the syslog yesterday are now gone.  No errors at all. Either the change in drive positions so all drives are now on the mobo or I had a loose cable that I tightened resolved the issue. It was the mobo connected drives that were reporting the funny errors. I researched those errors and there was lot of dated (2007-2008) info on exactly that error syndrome.  Consenseus opinion seemed to be that it was a bug in some linux distros. 

 

The bottom line is that the drives appear to be rock solid.  Drive temp of a spinning drive that is reading/writing runs between 35 and 37 degrees C. I assume it will be lower when the server is moved to its home which is getting special A/C. The errors I was seeing are in a post regarding parity on this forum section. 

 

After this disk is finished I will preclear the last disk and then include one of them in the array and keep one as a standby.

As an update to the first post in this thread:

 

The preclear finished after 24 hours 33 minutes.  This was a full 1 cycle preclear (pre-read, clear, post-read) on a Hitachi 2 TB 7200 RPM drive connected to a mobo (Gigabyte GA-MA785GM-US2H) Sata port via a reverse SAS-SATA breakout cable.  At the start of the pre-read cycle transfer rate was in the 120 mb/s range and at the end of the pre-read it was in the  64 mb/s range.  I am assuming that is due to disk geometry (Anyone know?).  The following messages appeared at the end of the preclear and I would appreciate it if someone could explain them to me:

 

Smart error count differences detected after preclear

Note, some raw values may change, but not an indication of a problem 19,20c19,20

Offline Data collection status: 0x80 Offline data collection activity was never started

Offline data collection status: 0x84 offline data collection activity was suspended by an interrupting command from the host

Interesting, my 1.5 TB WD EARS (5400 rpm) also took about 24 hours for one pass of preclear.  Interesting parallel.

 

I do believe the slowdown is due to disk geometry, and it completely normal.

well, with my SATA I mobo it has already taken 27 hours simultaneously preclearing 2 Hitachi hdds (2Tb , 7200, each same as barrygordon's) and it's only at 80% of post-read...

As an update to the first post in this thread:

 

The preclear finished after 24 hours 33 minutes.  This was a full 1 cycle preclear (pre-read, clear, post-read) on a Hitachi 2 TB 7200 RPM drive connected to a mobo (Gigabyte GA-MA785GM-US2H) Sata port via a reverse SAS-SATA breakout cable.  At the start of the pre-read cycle transfer rate was in the 120 mb/s range and at the end of the pre-read it was in the  64 mb/s range.  I am assuming that is due to disk geometry (Anyone know?).  The following messages appeared at the end of the preclear and I would appreciate it if someone could explain them to me:

 

Smart error count differences detected after preclear

Note, some raw values may change, but not an indication of a problem 19,20c19,20

Offline Data collection status: 0x80 Offline data collection activity was never started

Offline data collection status: 0x84 offline data collection activity was suspended by an interrupting command from the host

 

Yes, this is due to geometry.  Reads and writes are faster on the edge of the platter as the disk is physically moving faster than in the middle at the spindle.  I have run 5 preclear cycles on two of these drives and they always start out fast, around 125 Mb/s or so, but by the end they are down in the 60s.  By comparison, the 5 cycle preclear of my 5400 Samsung started around 105 Mb/s and ended at 48 or so.  FYI, there is a preclear thread that we all post our results to that might help you interpret your results.

well, with my SATA I mobo it has already taken 27 hours simultaneously preclearing 2 Hitachi hdds (2Tb , 7200, each same as barrygordon's) and it's only at 80% of post-read...

Sounds about right.  Doing 2 at the same time takes longer than doing only 1.

well, with my SATA I mobo it has already taken 27 hours simultaneously preclearing 2 Hitachi hdds (2Tb , 7200, each same as barrygordon's) and it's only at 80% of post-read...

Sounds about right.  Doing 2 at the same time takes longer than doing only 1.

 

are you sure? i thought that each sata port had its own bandwidth?

well, with my SATA I mobo it has already taken 27 hours simultaneously preclearing 2 Hitachi hdds (2Tb , 7200, each same as barrygordon's) and it's only at 80% of post-read...

Sounds about right.  Doing 2 at the same time takes longer than doing only 1.

 

are you sure? i thought that each sata port had its own bandwidth?

I guess I should qualify that then.

 

Yes, each port has its own bandwidth on its cable, but you are context switching between two processes.  That takes time.  You have twice as many blocks (potentially) in the buffer cache.    That may take longer to manage between the two processes.  (typically, we will constantly be re-using the least-recently-accessed blocks for the next I/O operation.  Unless you have a very large amount of RAM, there is no way all 2TB of each disk can be buffered. There are two processes needing to coordinate their access to the buffer cache.  Odds are high they use locks and semaphores to keep their accesses from clobbering each other. )

 

Now, if you have a dual core CPU, it might not be much different, and a disk controller that can handle multiple concurrent requests, it might not be any different at all, but, in my own experience, running multiple processes will cause them each to run a bit longer.

 

Joe L.

But your probably still faster doing 2 at once then one after the other..

But your probably still faster doing 2 at once then one after the other..

Yes, absolutely... way faster for most people.

But your probably still faster doing 2 at once then one after the other..

Of course. Run them one after the other, will take 10 hours each. Run them together, will take 11 hours total compared to 20.

 

Thanx for the info Joe

Although since it took 30 hours for my 2Tb Hitachi (instead of 24hours) i guess that the slow mobo is to blame a bit too...

That is much better than two spearate clears at 24 hrs apiece.

That is much better than two spearate clears at 24 hrs apiece.

 

That's true, although it depends on whether you are in a hurry to put a disk in use...

if yorue in a hurry just clear it, dont pre clear it. use a 2nd pc and free unraid if you have to.

if yorue in a hurry just clear it, dont pre clear it. use a 2nd pc and free unraid if you have to.

If you just "clear" it on another unRAID instance, you'll install it, but unRAID will then have to take many hours to calculate parity.  You'll be without parity protection until it completes.

if yorue in a hurry just clear it, dont pre clear it. use a 2nd pc and free unraid if you have to.

If you just "clear" it on another unRAID instance, you'll install it, but unRAID will then have to take many hours to calculate parity.  You'll be without parity protection until it completes.

 

well, since preclearing helps with locating problematic sectors, i think it is worth wasting the respective time

But you are right, if you are in a hurry you can avoid preclearing

 

+the parity thing that joe just mentioned