Seagate’s first shingled hard drives now shipping: 8TB for just $260

[SSD]

In general I would not recommend running more than 3 or 4 preclears at once. With some experimenting you might find you can do more if you have plenty of memory, relatively fast CPU, and your system is able to handle the high I/O requirements. There is no "simultaneous preclear" competition that I've heard of. At 6 you might be the winner if you went to completion.

 

Doing preclears while parity check is running is certainly possible, but I have never done it. I guess I am not very adventuresome.

 

I posted a faster preclear that you may have seen. The preread and zeroing are exactly the same, but the faster preclear speeds up the post read/verify stage by about 50%. So that is about a 1/6 reduction in a full cycle (so for a 24 hour cycle, it would run in roughly 18 hours).

 

I don't do multi-cycle preclears, but I believe the way they work is to only preread once - so for three cycles the operations would be "preread/zero/postread/zero/postread/zero/postread." If this is right, the postread/verify would be a bigger savings, as you are doing 3 of them out of the 7 full disk activities. So it would save close to half of the time for a 3 cycle operation.

[WeeboTech]

What are the failures?

[pras1011]

Two of the drives that were on the sas2lp failed to complete. Mbr could not completed or something. I checked all the drives smart and everything looks ok.

[garycase]

... So that is about a 1/6 reduction in a full cycle (so for a 24 hour cycle, it would run in roughly 18 hours).

 

Might want to check your math :)

 

[1/6th of 24 = 4;  24 - 4 =20  ... not 18.    Although I think you actually miscalculated the 1/6th ... since if you save 50% of the post-read time, which is typically about half of the total pre-clear cycle, then you've actually saved about 1/4th of the total time => which, ironically, happens to be a 6 hour savings in a 24 hour cycle  ]

 

 

[pras1011]

Anyway I started again. My server is running 2 preclears and the gaming pc is running 6 preclears. Server is refreshing the speed at 10-11 secs and pc is refreshing at 1-2 secs.

 

I am doing 2 cycles instead of 3 as all this nonsense has taken to long!!!

 

I have noticed that v6 recognises the hdds as non 4k but v5 does. Strange!

[johnsanc]

What are the key factors into how fast a parity sync will go with one of these drives?

 

My parity sync is currently very steady at 42 MB/s at about 1TB position now. Interestingly this is almost identical to danioj's transfer speed test.

 

My setup is in my signature. And my drives are:

1x  ST4000DM000-1F2168

1x  WDC_WD40EFRX-68WT0N0

1x  WDC_WD20EZRX-00DC0B0

6x  WDC_WD20EARS-00MVWB0

[pras1011]

This one has failed again.

 

Whats wrong in the syslog?

New_Text_Document.zip

[SSD]

... So that is about a 1/6 reduction in a full cycle (so for a 24 hour cycle, it would run in roughly 18 hours).

 

Might want to check your math :)

 

[1/6th of 24 = 4;  24 - 4 =20  ... not 18.    Although I think you actually miscalculated the 1/6th ... since if you save 50% of the post-read time, which is typically about half of the total pre-clear cycle, then you've actually saved about 1/4th of the total time => which, ironically, happens to be a 6 hour savings in a 24 hour cycle  ]

 

The post read takes about twice as long as the preread and zeroing steps. So saving half off of the post read means you are now saving 1 parts in 4 = 1/4 not 1/6.

 

In the second example, with preread, zero, postread, zero, postread, zero, postread, you are saving 3 in 10, a bit more than 1/3.

 

This is assuming that all three operations take about the same time with the fast preclear, which is not exactly true and varied based on server.

[pras1011]

Wasn't there an issue with unraid that the 32 bit version v5 doesn't use memory properly if 4gb plus is installed or something? I have installed v6 on my 8gb ram gaming pc and it's absolutely flying at about 190MB/s on zeroing 6 8tb drives.

 

Maybe that's the reason why preclear on the server failed?

[BRiT]

The v5 32bit OS will use the additionally memory, it's just that it's not as efficient in it's use since it has to use PAE instead of addressing it natively. There were some reports that the drive buffers wouldn't be as efficient with large file copies on 32bit v5 as well. There's also the contention with low memory on 32bit OS that doesn't happen on 64bit OS.

[pras1011]

It has taken about 70 hours to get through one cycle of preclear (using the older preclear), preclearing 6 x 8tb hdd through my gaming pc.

 

Cycle 2 has started.

 

All the smart reports are clean apart from one. See attached.

 

I think that disk is fine though but what do you lot think?

New_Text_Document_3.txt

[garycase]

It has taken about 70 hours to get through one cycle of preclear (using the older preclear), preclearing 6 x 8tb hdd through my gaming pc.

 

Cycle 2 has started.

 

All the smart reports are clean apart from one. See attached.

 

I think that disk is fine though but what do you lot think?

 

The 3 errors were all at the same sector and were very close together.  May have been a "glitch";  may have been an actual problem => but in any event didn't result in an actual failure, so it's probably okay.    I'd look carefully at the next two cycles and see if these continue ... if so, I'd RMA that drive.  Otherwise it's probably fine.

 

[opentoe]

Is there a pre-clear for Windows? I'm all filled up and don't have any available slots and want to keep my box up. I can use HD Tune to "burn-in" the drive and just make sure every sector is good, but unraid will still need to format it, which is no big deal.

 

 

 

[trurl]

Is there a pre-clear for Windows? I'm all filled up and don't have any available slots and want to keep my box up. I can use HD Tune to "burn-in" the drive and just make sure every sector is good, but unraid will still need to format it, which is no big deal.

Formatting is no big deal, because all formatting means is creating an empty file system. The issue is having a disk clear (completely zeroed) so it matches existing parity before adding it to the array.

 

It sounds like you don't plan to add it to the array, though, since you say you don't have any slots. Will this be a replacement for another disk? Rebuilding an existing disk to a larger disk does not require a clear disk, since the rebuild will already match parity. It is still a good idea to test any new disk though. It can be tested using another program on another computer, as you suggest.

 

Since you did say format, though, perhaps I don't understand what you intend. If a rebuild is indeed what you have in mind, note that formatting a disk is NEVER part of the rebuild process.

[c3]

It has taken about 70 hours to get through one cycle of preclear (using the older preclear), preclearing 6 x 8tb hdd through my gaming pc.

 

Cycle 2 has started.

 

All the smart reports are clean apart from one. See attached.

 

I think that disk is fine though but what do you lot think?

 

Given the power-on hours, can you figure that was during the pre-read phase? With no reallocate, it looks like an initial read failed. That's one area of preclear weakness, it assumes the drive is readable.

[garycase]

Agree with c3 => and if that is indeed the case the next 2 cycles will be fine.

 

[garycase]

... Is there a pre-clear for Windows? I'm all filled up and don't have any available slots and want to keep my box up. I can use HD Tune to "burn-in" the drive and just make sure every sector is good

 

There's no pre-clear script for Windows; but you can certainly test the drive in Windows -- using HD Tune;  WD's Data Lifeguard;  Seagate's SeaTools; or any of several other disk utilities.    However, even if you write zeroes to the drive as a final step, it will won't be "zeroed" to UnRAID's standard, which includes a special "pre-clear signature".

 

 

... unraid will still need to format it, which is no big deal.

 

I assume (as mentioned above) that you've going to REPLACE a drive with this drive ... in which case it will NOT require formatting.    However, if you are planning to ADD it (perhaps by adding an additional controller or by restructuring your array before this drive is ready), then it will NOT just require formatting => it will require zeroing the entire drive -- THEN you can add it; and THEN it will need to be formatted (a quick process).    Note, however, that during the zeroing of the drive by UnRAID your array is NOT available.  [Eliminating this downtime is probably the biggest advantage of the pre-clear script]

 

 

 

[pras1011]

Those errors indeed happened on the pre read stage of cycle 1.

 

The pre read on cycle 2 has gone through without any problems.

 

Can I just do 2 cycles instead of 3?

[SSD]

... Is there a pre-clear for Windows? I'm all filled up and don't have any available slots and want to keep my box up. I can use HD Tune to "burn-in" the drive and just make sure every sector is good

 

There's no pre-clear script for Windows; but you can certainly test the drive in Windows -- using HD Tune;  WD's Data Lifeguard;  Seagate's SeaTools; or any of several other disk utilities.    However, even if you write zeroes to the drive as a final step, it will won't be "zeroed" to UnRAID's standard, which includes a special "pre-clear signature".

 

I have sometimes wondered if the drive was zeroed with no preclear signiature, if you could do a new config, reassign all old drives + 1 new zeroed drive (with no signative), and do a trust parity, if you would then accomplished the same thing with no signature.

[trurl]

I have sometimes wondered if the drive was zeroed with no preclear signiature, if you could do a new config, reassign all old drives + 1 new zeroed drive (with no signative), and do a trust parity, if you would then accomplished the same thing with no signature.

Interesting question. Is the signature omitted from parity, or does adding a drive update parity with the signature? Probably only Tom knows.

[SSD]

I have sometimes wondered if the drive was zeroed with no preclear signiature, if you could do a new config, reassign all old drives + 1 new zeroed drive (with no signative), and do a trust parity, if you would then accomplished the same thing with no signature.

Interesting question. Is the signature omitted from parity, or does adding a drive update parity with the signature? Probably only Tom knows.

 

unRAID sees the signature only as the disk is added. It then creates the partition. The only reason this might not work is that unRAID ALWAYS creates a partition as a disk is added to the array. In this case you would be slamming a disk with no partition into the array. This could confuse unRAID and force you to manually create that partition before the disk is usable. But I expect if unRAID sees an partitioned disk, it would partition it. I may have to try it sometime.

[pras1011]

This is strange. Cycle 1 took 70 hours. I am on cycle 2, its only been 17 hours (87 hours in total including cycle 1 hours) and its currently at the post read stage.

 

Cycle 2 seems very very fast!!

 

WTF?

[SSD]

This is strange. Cycle 1 took 70 hours. I am on cycle 2 post read and its only been 17 hours (87 hours in total including cycle 1 hours).

 

Are you using the fast preclear? Did you copy the readvz file correctly?

[pras1011]

I am using the the standard preclear.

syslog-2015-05-06.txt

[c3]

This is strange. Cycle 1 took 70 hours. I am on cycle 2 post read and its only been 17 hours (87 hours in total including cycle 1 hours).

 

Cycle 2 seems very very fast!!

 

WTF?

17 hours is expected range for read cycle.