Preclear at 9MB/s on Samsung 2TB

Hi there,

I removed a 2TB drive (SAMSUNG_HD203WI_S1UYJDWZ609551) from my desktop PC over the weekend, and stuck it in my (new) unRAID box (HP MicroServer gen8 w/ a Celeron G1610T @ 2.30GHz and 16GB ECC RAM). The disk is a few years old.

I started the preclear from the GUI before I went out on Saturday, and it wasn't until today that I saw most speeds for preclear were up near or above 100MB/s, while mine has been crawling along at about 10MB/s since I started the process. It's currently at 88% Pre-read (after 49h:35m), so I think I'll let it finish, but does anyone have any idea why this is so much slower?

I've attached my syslog.

Also, FWIW, I also had slow speeds when building the parity disk, which I posted about here: https://lime-technology.com/forum/index.php?topic=50194.msg481766

I have a 4TB Seagate NAS as parity (ST4000VN000-2AH166_WDH00MXQ sdb), and another identical disk as storage (ST4000VN0002AH166_WDH00GP7 - sdc).

I didn't use the preclear script/GUI to preclear these disks, so can't compare how fast they were.

Your help and insight is appreciated, thanks.

percy-syslog-20160711-1436.zip

And here are my anonymised diagnostics if they're of help, too.

percy-diagnostics-20160711-1459.zip

From the SMART report for that Samsung disk:

197 Current_Pending_Sector  0x0032  100  100  000    Old_age  Always      -      3

Pending sectors are bad news.

I suggest you do a long SMART self test on it and post your results.

From the SMART report for that Samsung disk:

 

197 Current_Pending_Sector  0x0032  100  100  000    Old_age  Always      -      3

 

Pending sectors are bad news.

 

I suggest you do a long SMART self test on it and post your results.

 

Will do mate. It's now zeroing @ 97 MB/s (58:50:42), so will let it finish before I do.

Notably, the SMART error has now cleared itself:

2016-07-11 23:24 unRAID device sdd SMART message [197] Notice [PERCY] - current pending sector returned to normal value SAMSUNG_HD203WI_S1UYJDWZ609551 (sdd) normal

And from the SMART info screen:

197 Current pending sector 0x0032 252 100 000 Old age Always Never 0

From the SMART report for that Samsung disk:

 

197 Current_Pending_Sector  0x0032  100  100  000    Old_age  Always      -      3

 

Pending sectors are bad news.

 

I suggest you do a long SMART self test on it and post your results.

 

 

Will do mate. It's now zeroing @ 97 MB/s (58:50:42), so will let it finish before I do.

Notably, the SMART error has now cleared itself:

2016-07-11 23:24 unRAID device sdd SMART message [197] Notice [PERCY] - current pending sector returned to normal value SAMSUNG_HD203WI_S1UYJDWZ609551 (sdd) normal

And from the SMART info screen:

197 Current pending sector 0x0032 252 100 000 Old age Always Never 0

If that is the case, then Attribute #5  (Reallocated sector count) should now be 3!

I would suggest that you repeat the preclear operation until the #197 remains zero and #5 does not change from the count at the end of the previous cycle.  If that does not happen by the end of the third cycle, I (personally) would replace the disk or RMA it (if still in warranty)!

Pending sectors don't always get reallocated, sometimes they work when the drive tests them again. A minor power blip can cause a pending sector that is really ok, but at the instant the drive tried it the first time, it failed because of the power fluctuation.

In any case, you definitely want stability. I'd shoot for at least 2 consecutive preclears where the critical attributes stayed the same.

If that is the case, then Attribute #5  (Reallocated sector count) should now be 3!

 

I would suggest that you repeat the preclear operation until the #197 remains zero and #5 does not change from the count at the end of the previous cycle.  If that does not happen by the end of the third cycle, I (personally) would replace the disk or RMA it (if still in warranty)!

I initially intended to run the preclear three times, as per the recommendation, however given it takes this long... maybe just twice.

It's now up to the post-read stage, and it's doing even worse:

Preclear in progress... Post-Read (1 of 3). 3% @ 5 MB/s (67:24:28)

Curiously, in the preclear log it says it's reading at 10.5MB

Pending sectors don't always get reallocated, sometimes they work when the drive tests them again. A minor power blip can cause a pending sector that is really ok, but at the instant the drive tried it the first time, it failed because of the power fluctuation.

 

In any case, you definitely want stability. I'd shoot for at least 2 consecutive preclears where the critical attributes stayed the same.

Seems you're right:

5 Reallocated sector count 0x0033 252 252 010 Pre-fail Always Never 0

Hrmmm just found this:

https://lime-technology.com/forum/index.php?topic=24992.0

Problem fixed by turning on write caching in the BIOS. Hope this info will be useful to anyone else with this server.  Now getting preclear speed of 125MB/s instead of 13MB/s

Thoughts on if this applies to me?

I'm 72 hours in to my preclear, while 7% (@ 5 MB/s) through the Post-Read stage, so I'm unsure if I should cancel it now to check, or wait the next three or so days for this step to finish - I odn't want to have to start the process from scratch again...

Do I wait? Do I start again?

I have one more observation.  The GUI has been know to radically effect the speed of the parity check for many systems because the GUI interrogates the drives when it refreshes the display as I recall.  So if you watching it all the time that could be impacting what you are seeing.  Plus, I am not really certain how either of those two speeds are calculated and what they truly signify.  At best, they are just an indicator of how things are going...

But I will say that beta preclear plugin seemed to be slower than the preclear plugin (using Joe L's preclear script) but I will admit that I never ran a full cycle on the beta script.  But, in my experience,  the preread and post read have always been slower than the writing of zeros to the drive even when just the script was run on the console. 

I have one more observation.  The GUI has been know to radically effect the speed of the parity check for many systems because the GUI interrogates the drives when it refreshes the display as I recall.  So if you watching it all the time that could be impacting what you are seeing.  Plus, I am not really certain how either of those two speeds are calculated and what they truly signify.  At best, they are just an indicator of how things are going...

 

But I will say that beta preclear plugin seemed to be slower than the preclear plugin (using Joe L's preclear script) but I will admit that I never ran a full cycle on the beta script.  But, in my experience,  the preread and post read have always been slower than the writing of zeros to the drive even when just the script was run on the console.

Good to know, thanks. FWIW, I only periodically open the GUI to check on the progress.

Ok, so I cancelled the preclear around 75 hours in, and rebooted to BIOS to enable write cache.

The preclear speed is now up at ~74MB/s

I know it's naughty doing this without a UPS, so I shall buy one very soon.

If I don't use it to power my main PC, I can get away with a £65 one (CyberPower VALUE 600EILCD, which I believe is supported natively in unRAID).

FYI, the Samsung HD203WI has an issue with slow parity check/syncs and disks rebuilds when used on a AHCI controller, so you're not surprised, for more info see this thread:

https://lime-technology.com/forum/index.php?topic=42384.0

FYI, the Samsung HD203WI has an issue with slow parity check/syncs and disks rebuilds when used on a AHCI controller, so you're not surprised, for more info see this thread:

 

https://lime-technology.com/forum/index.php?topic=42384.0

Oh wow. I'm really glad you posted that, thanks.