Mix of varying quality/age drives how do I choose what to use

[alexdodd]

I have a mix of old 500gb, 1tb and 2tb drives.  I can't use them all, and I want to use the best ones in my system. I know whatever I do is going to take a long time.  Which means I want to do it right once

 

What is the best accepted way on the forum now days for doing this, either on unraid, or on my windows box...

Just taken 6 months off because I moved house, changed job and had a lot going on, thought this weekend, I would just fire up the server and do a bit of work, so far it's taken the whole weekend!!

Final hurdle is I have loads of drives that are 3-7 years old, with power on times and cycles that are not linked to age, also it seems varying bad secor counts, I just put one in my unraid box and it got me worried. 

So i'd rather just do a whole load of stress testing and whatever now and get a better idea, before I start using parity and rebuilding as that stress test!  Everything works fine right now, and i'm happy with how its working, and i'd rather just add a few drives in there to up the space since I have them.

This all leads me to preclear, which was something last time I was here, and now its not, and i'm not trying to start arguments etc. Either way its not been updated, and I've seen version that "works", but its a hack and i dont know why it works.

As i said I don't mind what I use, as long as thats the consensus, I really dont want to waste the next few days ?

 

 

[alexdodd]

For clarity, I do have that hacked version of preclear working away on a 1tb drive, but its grinding down to a halt at 13MB/s now on only 23%, and the errors are rolling in thick and fast.  So I presume that drive is toast (things probably got tossed around way more than they should have in that move?), or the preclear script is broken, how do I know or trust?




 

 

Just after posting this, the preclear script has paused too, so maybe its the drive I guess?

 

Edited March 18, 2018 by alexdodd

[pwm]

I think you should use your old drives for backup. Having parity doesn't make it safe to use bad drives - it just gives you more options to waste time when the bad drives starts to give problems and you then depend on the other drives to work flawlessly while you perform your rebuilds.

 

Look at 2-3 new drives for your storage server - you will be much happier with drives you can trust. And larger drives will consume less power in relation to capacity.

[alexdodd]

3 minutes ago, pwm said:

I think you should use your old drives for backup. Having parity doesn't make it safe to use bad drives - it just gives you more options to waste time when the bad drives starts to give problems and you then depend on the other drives to work flawlessly while you perform your rebuilds.

 

Look at 2-3 new drives for your storage server - you will be much happier with drives you can trust. And larger drives will consume less power in relation to capacity.

 

 

Whilst I appreciate that, i'm on a budget (just moved ;)), I have lots of drives, some aren't used much and others more so.  Some drives have been upgraded when I got an SSD, or because I needed more room, or been decommissioned from PC's, so *some* of these are fine.  Granted the one above appears not, but that is also because the moving process wasnt kind to a lot of stuff.
 

So buying new is not answer here, when I can test 15 drives I've collected (including the duff one above) and not thrown away.  It's about choosing the best and loosing the rest.  I would like to know what the best method is since its a time consuming process!

 

[JorgeB]

Run an extended SMART test on all and when done post all the SMART reports.

[pwm]

If you run unRAID with n data drives and 1 parity you can only afford to lose one (1) drive.

And if you need to repair, you can't afford a single failed read from any of the remaining drives.

So bad drives are just not an option, even if you are on a budget.

[Guzzi]

If you really want to go with the old drives , I'd suggest you first test smart values for all drives, throw away the drives that fail or show many errors and then run preclear on the error free ones. Finally you should have a set of drives that should be good to go.

 

[alexdodd]

3 minutes ago, pwm said:

If you run unRAID with n data drives and 1 parity you can only afford to lose one (1) drive.

And if you need to repair, you can't afford a single failed read from any of the remaining drives.

So bad drives are just not an option, even if you are on a budget.

Agreed. 

All the drives won't be in use on unraid in parity, they will be ready for swapping because I have too many, but I want to know now which ones to keep, or have a vague idea before the time arrives.

 One will become a second parity drive.  Some may just be mounted to use separate over the network outside of unraid.

 

 

Edited March 18, 2018 by alexdodd

[Guzzi]

And as mentioned by others - do never ever use bad drives - it is not worth the trouble you'll get - so if smart reports issues, you should really dispose those drives.

[alexdodd]

 

2 minutes ago, Guzzi said:

If you really want to go with the old drives , I'd suggest you first test smart values for all drives, throw away the drives that fail or show many errors and then run preclear on the error free ones. Finally you should have a set of drives that should be good to go.

 

Preclear seems to be showing the first drive i put in it to be a lemon, which might very well be the case but because its unsupported I came here to check there wasnt a better way, as i said, i dont want to waste the week to find out preclear isnt really working properly....

 

 

6 minutes ago, johnnie.black said:

Run an extended SMART test on all and when done post all the SMART reports.

Thanks guys, an extended smart test should take agggges, so as long as that's enough, I thought preclear was "better" because it was some sort of write read test, but maybe thats what an extended smart test is?

[alexdodd]

3 minutes ago, Guzzi said:

And as mentioned by others - do never ever use bad drives - it is not worth the trouble you'll get - so if smart reports issues, you should really dispose those drives.

 

That's what this exercise it all about, these drives have been pilling up for years on my "test" list that never ever happens.  I want a once and for all test and throw method, if smart extended is the way, then I shall do that

[JorgeB]

Just now, alexdodd said:

Thanks guys, an extended smart test should take agggges, so as long as that's enough, I thought preclear was "better" because it was some sort of write read test, but maybe thats what an extended smart test is?

 

Extended SMART test takes the same time as a write only preclear, and much less time than 1 preclear cycle.

[pwm]

Best is preclear and then an extended SMART to read back all sectors.

 

Remember that the drive do not use any feedback when doing writes - if the drive just believes it has found the correct track and sector location and thinks the head is at the correct flying height it will perform the write and trust the result. The extended SMART will read every single sector and verify that they can be read back correctly.

[alexdodd]

1 minute ago, johnnie.black said:

 

Extended SMART test takes the same time as a write only preclear, and much less time than 1 preclear cycle.

 

Coolio, so that gets me back to is there another method that is more reliable?  Even outside of unraid?  I'm wanting water right results, regardless of time, do it once, throw away the lemons.

[alexdodd]

HDSentinal does exactly what i'm after, purchased, and can queue up loads of drives, give stats/data, and everything after it's finished.  In case anyone else looking for a solution to my specific problem.

Surface tests can be desructive with read - write - read etc... too, meaning I get a real overall view of everything.  

This may not be for everyone but as i said, I want to set this going and leave it,and come back with results.  The software has already paid for itself in this case.  Shame preclear wasn't working quite right.

 

One hard disk is already looking totally paperweight worthy.  The rest seem great

[alexdodd]

Reports.zip In case anyone is interested, although i doubt it really.  These are some i remembered to save as i went along, every drive was reinitialised.  I still have to test 8 2tb SAS drives, but the card is being used in my server, so I'm unwilling to turn it all off right now, or play around using it to check over those.  I really like the software :).

 

So the real bulk are still outstanding, and as predicted, a lot just need to be thrown away.  Some golden 500gb gems though, of which I have no idea yet what I can actually do with.  But they're in perfect condition (for age) and very little spin time as predicted.

 

Cheers

[alexdodd]

Here's the big test, I'm reinitializing and surface testing them all, I have noticed that Pre Clear got the all clear for 6.5.  but since i've tested all the other drives with HD Sentinel and since it gives a html report, and i had to mess around earlier this week to switch the SAS card around, i'm now sticking with HD sentinel.

 

I'm sure pre clear will have its uses in the future

 

 

[alexdodd]

Final report, pretty happy I didn't just throw them all away as originally suggested, I have some total gems

2 with only 25 days power on time with 5.6 and 23tb of writes resp, including my surface disk reinitialise.
3 with 238 days with 5, 5 and 16tb writes
3 with 2038 days  but all only 5.6tb writes

All with 100% passes on everything you could measure on a full surface reinitialise and test.

Full details here if anyone really (unlikely) cares:
https://drive.google.com/open?id=1TPqtH19f9tEQ50NiLDHz_-v6fmvOLV9F

Example:

[JonathanM]

As long as your electric costs are low, and you already have plenty of ports and physical space, then many small drives are fine. When you need absolutely the lowest power consumption and heat production, or are short on ports, then fewer larger drives are the answer.

 

Also, only run as many drives as you need to keep the data you want. Don't just install all the healthy drives just because. Add drives as you need to add capacity. That is the beauty of unraid, it's easy to add drives whenever you want. It's harder to reduce your spindle count, so keep the number of drives as small as possible.

[alexdodd]

On 03/04/2018 at 2:18 AM, jonathanm said:

As long as your electric costs are low, and you already have plenty of ports and physical space, then many small drives are fine. When you need absolutely the lowest power consumption and heat production, or are short on ports, then fewer larger drives are the answer.

 

Also, only run as many drives as you need to keep the data you want. Don't just install all the healthy drives just because. Add drives as you need to add capacity. That is the beauty of unraid, it's easy to add drives whenever you want. It's harder to reduce your spindle count, so keep the number of drives as small as possible.

Certainly, now these are tested, and confirmed good, they will be installed in the server, but 6 out of 8 will be currently left sitting as hot spares for space extension and/or failover