Cheap Hitachi 2TB Drives (Data Center Recycle) $30! Free Shipping!

ttabbal

Active Member
Mar 10, 2016
743
199
43
43
Do you mind sharing your stress test methodology?
It's not all that complicated.

Look at the SMART data and check the number of reallocated sectors etc.. If you see bad info here, they fail.

I run badblocks on them, letting it do the full 4 write/read passes. This takes days. Let it run. If badblocks reports errors, fail.

Check SMART again, if the data still looks good, run a SMART long test on them. This takes about 6 hours for these drives.

If SMART still looks good, you're probably good to go. I like to also create a ZFS mirror across all the drives, fill it up, let it sit, scrub, sit, scrub, and maybe delete it all and fill it again, depending on how trusting I'm not feeling about drives that day. :) The ZFS test is best done on the server they will live in. Mostly to test the cables, controllers, backplane, etc. etc..
 

josh

Active Member
Oct 21, 2013
310
87
28
It's not all that complicated.

Look at the SMART data and check the number of reallocated sectors etc.. If you see bad info here, they fail.

I run badblocks on them, letting it do the full 4 write/read passes. This takes days. Let it run. If badblocks reports errors, fail.

Check SMART again, if the data still looks good, run a SMART long test on them. This takes about 6 hours for these drives.

If SMART still looks good, you're probably good to go. I like to also create a ZFS mirror across all the drives, fill it up, let it sit, scrub, sit, scrub, and maybe delete it all and fill it again, depending on how trusting I'm not feeling about drives that day. :) The ZFS test is best done on the server they will live in. Mostly to test the cables, controllers, backplane, etc. etc..
Do you test with ZLOG and read caches?
 

josh

Active Member
Oct 21, 2013
310
87
28
It's not all that complicated.

Look at the SMART data and check the number of reallocated sectors etc.. If you see bad info here, they fail.

I run badblocks on them, letting it do the full 4 write/read passes. This takes days. Let it run. If badblocks reports errors, fail.

Check SMART again, if the data still looks good, run a SMART long test on them. This takes about 6 hours for these drives.

If SMART still looks good, you're probably good to go. I like to also create a ZFS mirror across all the drives, fill it up, let it sit, scrub, sit, scrub, and maybe delete it all and fill it again, depending on how trusting I'm not feeling about drives that day. :) The ZFS test is best done on the server they will live in. Mostly to test the cables, controllers, backplane, etc. etc..
Is there a way to queue testing of all the drives? What block size do you use?
 

josh

Active Member
Oct 21, 2013
310
87
28
Alright I just got half of the drives. I plug the first two in and launch HD Sentinel Trial and am immediately greeted with this. Looking forward to the rest!

There are 120 bad sectors on the disk surface. The contents of these sectors were moved to the spare area.
Based on the number of remapping operations, the health of the disk was decreased in different steps.
It is recommended to examine the log of the disk regularly. All new problems found will be logged there.

It is recommended to backup immediately to prevent data loss.
 

ttabbal

Active Member
Mar 10, 2016
743
199
43
43
All the tests I do are command line. So I just open however many sessions I need with screen and run them once per drive.
 

josh

Active Member
Oct 21, 2013
310
87
28
All the tests I do are command line. So I just open however many sessions I need with screen and run them once per drive.
You test them in your JBOD? Linux?

So I've had 2/4 drives with bad sectors right out of the box
 

xnoodle

Active Member
Jan 4, 2011
259
48
28
With my personal arrays, this was my process:

Plug them in to a desktop running Linux.
Generate a copy of the smart data and store it.
Run a long smart test and check afterwards.
Run a full badblocks pass (-wsv flags IIRC).
Generate another copy of the smart data and compare it.

I would run 4-6 at a time.

Any drives that took substantially longer (easy to see when you run them in parallel, saw one that would only write at 10megabytes/second), had/incremented questionable SMART values, or outright failed badblocks were marked bad.
 

ttabbal

Active Member
Mar 10, 2016
743
199
43
43
My process is similar to xnoodle. I used my desktop and my backup server to process 11 drives at once. It would have been 12, but I had one DOA.

Even brand new drives sometimes have remapped sectors. That's not a big deal. What matters is if that count increases after badblocks or if it runs out of spare sectors and has to report errors to the OS.
 

dragonme

Active Member
Apr 12, 2016
282
28
28
josh.. while ttabbal is correct.. even new drives can have bad sectors and in and of itself does not mean a failing drive if they stay stable after a couple rounds of badblocks.. this seller will refund or exchange for that... just ping them on eBay.. sometimes you will be given a refund and told not to send them back if you wish....

stand up sellers.. they are not hard drive experts and I dont think anyone there is running smart tests on them.. at best they just see if it passes or fails and as you know .. you have to go below threshold for the failure message...

that being said they have a great customer service guys.. I have now delt with 2 different people and both have been super great... and I only bought 10 drives.. not 5k... so no reason for them to kiss my ring either...
 

klammore

Member
Jun 17, 2016
77
15
8
40
received 5, 1 was bad out of the box. Stress testing the other 4 now, 9 hours into the tests and they seem to be going well so far.
 

josh

Active Member
Oct 21, 2013
310
87
28
My process is similar to xnoodle. I used my desktop and my backup server to process 11 drives at once. It would have been 12, but I had one DOA.

Even brand new drives sometimes have remapped sectors. That's not a big deal. What matters is if that count increases after badblocks or if it runs out of spare sectors and has to report errors to the OS.
Except when you see this with the first drive you plug in

 

josh

Active Member
Oct 21, 2013
310
87
28
Disk #3. I'd stay away from these drives if you're looking to store anything of importance

 

dragonme

Active Member
Apr 12, 2016
282
28
28
Disk #3. I'd stay away from these drives if you're looking to store anything of importance

josh.. um.. I dont use drive sentinel I use unbuntu live and smartctl and bad blocks raw.. so you have to translate what spooks you about

Status

GOOD 2058
DAMAGED 0
BAD 0

unless you are talking about runtime?

is that speed equivelent for the other drives as well or do you just have one that is slow?

it could be the type of test you are trying to run?

out of the 10 I got I have 2 that run a bit slower than the others.. but less than 10%

once again.. dont focus on what you got.. focus on the fact that once you have given the drives the once over with whatever process you want to check them with.. provided it doesnt take 180 days.. and ask the seller to replace or refund whichever drives are not performing...

again.. they took back a drive that had CRC errors.. the drive had 24000 hours on it and the CRC errors were logged at like 1000 hrs.. which was probably a loose cable IE not an issue.. but they replaced it no questions...

again... these drives have low hours compared to others I run.. and just because smart values are no longer perfect does not mean that the world will crash tomorrow... as drives age they will slow a bit.. and posssibly increase read, write seek time errors and decrease a bit in speed.. its mechanical after all ...

but if that drive is significantly slower than others.. just ask to have it replaced... I dont want to sound rude .. but stop whining...

now if you got a crap drive and the seller says you have to eat it.. by all means come back.. and bitch about the seller so we can avoid them in the future..

I bought a new honda once that had a bad fuel injector right from the start.. they replaced it.. doesnt mean I wont ever buy another honda... ?
 
  • Like
Reactions: knightcrusader

josh

Active Member
Oct 21, 2013
310
87
28
josh.. um.. I dont use drive sentinel I use unbuntu live and smartctl and bad blocks raw.. so you have to translate what spooks you about

Status

GOOD 2058
DAMAGED 0
BAD 0

unless you are talking about runtime?

is that speed equivelent for the other drives as well or do you just have one that is slow?

it could be the type of test you are trying to run?

out of the 10 I got I have 2 that run a bit slower than the others.. but less than 10%

once again.. dont focus on what you got.. focus on the fact that once you have given the drives the once over with whatever process you want to check them with.. provided it doesnt take 180 days.. and ask the seller to replace or refund whichever drives are not performing...

again.. they took back a drive that had CRC errors.. the drive had 24000 hours on it and the CRC errors were logged at like 1000 hrs.. which was probably a loose cable IE not an issue.. but they replaced it no questions...

again... these drives have low hours compared to others I run.. and just because smart values are no longer perfect does not mean that the world will crash tomorrow... as drives age they will slow a bit.. and posssibly increase read, write seek time errors and decrease a bit in speed.. its mechanical after all ...

but if that drive is significantly slower than others.. just ask to have it replaced... I dont want to sound rude .. but stop whining...

now if you got a crap drive and the seller says you have to eat it.. by all means come back.. and bitch about the seller so we can avoid them in the future..

I bought a new honda once that had a bad fuel injector right from the start.. they replaced it.. doesnt mean I wont ever buy another honda... ?
No the drives are mounted and therefore should have good access time to all sectors like this:



Explanation of what the dark blocks mean - "These are accessed slower":

 

dragonme

Active Member
Apr 12, 2016
282
28
28
No the drives are mounted and therefore should have good access time to all sectors like this:



Explanation of what the dark blocks mean - "These are accessed slower":

josh..

it it is materially slower than your other drives make note of the RW speeds and just RMA the drive... done...

again.. focus on the fact that

1 the price per TB rocks
and
2 the customer service is good so you will come out happy

3 you will then have enterprise drives that have very low hours on them for a 1/10 the price per TB of commercial garbage..

hitachi drives are great and I have my primary pool on their consumer drives and have over 47000 hours on that pool.. no issues...
 

klammore

Member
Jun 17, 2016
77
15
8
40
RMA'd my 1 dead drive, the 4 other passed fine. Super easy process.

3 of the good drives had 24-26k hours and the other good one has 11k hours, great deal in my book.
 

josh

Active Member
Oct 21, 2013
310
87
28
Yea the seller is responsive, the packing is good, the price is good. I'm not going to deny that. The only problem is the drives have been abused thoroughly and I'm just warning anyone not to store anything of importance on these drives or make sure you do some really extensive testing.
 

ttabbal

Active Member
Mar 10, 2016
743
199
43
43
I don't consider the drives to be "abused". They ARE used, my only complaint with them is that the seller calls them "refurbished". IMO, that label should only apply if the manufacturer actually fixed them up, not "I pulled them out of an array case and put them in anti-static bags".

I have consumer level drives with 2x the hours that I just pulled. Most of them work fine, but they are 1.5TB so it didn't make sense to keep them in the main system when these had lower hours and are "enterprise" drives. I'm not convinced the "enterprise" label is particularly meaningful, but we'll see. Everyone should thoroughly test every drive, brand new, used, or sitting on a shelf as a spare. Not doing so is asking for trouble. I've had brand new drives from every manufacturer fail to make it though burn-in in about the same ratio as I've seen posted in this thread for these. That's just life with spinning rust. If you care about your data, you burn-in all hardware, particularly drives, and use redundant storage configurations. And you have backups of anything important. Trusting any drive with the only copy of your data is ridiculous. Even my pr0n is on redundant storage and backed up, just sayin...
 

canta

Well-Known Member
Nov 26, 2014
1,028
198
63
39
well, not to stir up the situation:
used drive must be looked by
1) how many hours of up tomes
2) how many blocks had been written
3) smart status

As my experience, new drive ( not crap green/blue or low power drives :) always serves me well and never be disappointed.
some case was, packing was nasty.. that would damage on drive. Newegg was a bad seller when talking packaging of drives.

my best of the best is: Seagate 500G (OLD model), and Samsung Drives ( before Hard drive division got sold).
but as always, backup!!! before happens.

what is enterprise? can you trust ent drive with so many hours of uptimes?
my bible says checking how many block had been written, is the key since the platter would degrade when writing happened.

when I want to buy used drive, ( some seller said, refurbished, and I said you lie hehe), upttime shoud ble less than 1000 hours, or prefer 100-500 hours.

based on my experiences, used up drives has tendency to fail very soon regardless of enterprise or consumer ( not crap low power/green/blue/whatever). recovery is time consuming for me as today.
I am collecting Toshiba 3T, new or slightly used.. to expand my raidz2 to raidz3. just waiting till the deal shown up,
Just score 2 drives new :), Toshiba drives

please disagree with my experience, since everyone has his/her own too....

but... if this 2T drives dropt to $15, I would buy 10 for fun :D