New to FreeNAS, planning first build

[Woodward]

I'm excited to be putting together some good reliable storage here, finally. It will be for a home network, storing files for a couple of small businesses, as well as family music, movies, photos, etc.

Browsing this forum has been very helpful. I have nothing yet, but ordered and on its way here:
MB: Supermicro X9SCM-F
Case: Fractal Designs R4
Power supply: Seasonic SSR-360GP
Memory: Crucial 8GBx2 DDR3 PC3-14900 Unbuffered ECC 1.5V 1024Meg x 72

Got some good deals from Newegg and eBay. Total so far for all of the above only $385.00.
I still need to choose a CPU and hard disks. I don't yet have enough understanding of vdevs and zpools to decide about numbers and sizes of disks, and Z2 or Z3. I'm looking at WD Greens from Newegg right now as there is a special promo ending today. I can get five 3TB WD Greens for $424.95. Not sure whether to get the five and put in a 5-disk Z3 zpool in a single vdev, or get six and have a spare on the shelf, or get six and put them all in, filling the MBs SATA ports, and either use Z3 with compression or use Z2. Will appreciate any comments or suggestions.

 

[Fuganater]

Why not get NAS grade drives?

 

[Bidule0hm]

Because it's not mandatory, greens are good once you change the park timer with WDIDLE3.

RAID-Z3 with 5 or 6 drives is very overkill but if you're very very paranoid then why not. I'd recommend RAID-Z2 with 6 drives ;)

 

[Fuganater]

Green drives also run hotter than Red.

 

[Robert Trevellyan]

I'd recommend RAID-Z2 with 6 drives

+1, with compression (because always with compression)

 

[Nick2253]

Green drives also run hotter than Red.

That's simply not true. In fact, if there is any difference between them, the greens would likely run just lightly cooler than the reds because they are programmed to be more aggressive about parking the drive head.

From everything we can tell, Greens and Reds are identical hardware, with different software (the Red's extra cost is basically paying for an additional year of warranty). The only material software difference is the idle timer. This is confirmed by the WD spec sheets, tear downs, chip comparisons, etc. If anything the Reds are simply binned Greens, though there's no evidence of that either way.

 

[Robert Trevellyan]

The only material software difference is the idle timer

That's simply not true ;)

Reds support TLER, which makes a big difference or none, depending on how you want your FreeNAS to behave when a disk goes bad.

 

[Nick2253]

That's simply not true ;)

Reds support TLER, which makes a big difference or none, depending on how you want your FreeNAS to behave when a disk goes bad.

You got me :D

Honestly, from the perspective of a home NAS user, I would say that the idle timer is significantly more important than TLER. So I guess that hinges on what I meant by "materially". Cyberjock said it best:

TLER is just as important for software RAID as it is for hardware RAID. The real question is whether you can handle the potential problems that come with not having TLER. If you serve movies and stuff and you can handle it becoming almost unresponsive when a disk starts going bad, then you don't need TLER. If you're going to have wife aggro that is going to make WWII look like a fun ride at a park, then go for TLER.

I believe you can enable TLER on the greens as well, though my knowledge on that front is more than a few years out of date.

 

[Fuganater]

That's simply not true. In fact, if there is any difference between them, the greens would likely run just lightly cooler than the reds because they are programmed to be more aggressive about parking the drive head.

From everything we can tell, Greens and Reds are identical hardware, with different software (the Red's extra cost is basically paying for an additional year of warranty). The only material software difference is the idle timer. This is confirmed by the WD spec sheets, tear downs, chip comparisons, etc. If anything the Reds are simply binned Greens, though there's no evidence of that either way.

https://www.backblaze.com/blog/hard-drive-temperature-does-it-matter/

2C might not seem like much but when you put 20 of them in a chassis together, it can be a huge differance.

 

[Nick2253]

First off, their data show a 0.41C difference between the Reds and the Greens of comparable models, not 2C.

You do realize that the Backblaze data is really statistically unsound, right? For all we know, that's well within the margin of error (which should be published along with an average in order to give meaning to that average). And based on more rigorous data, like Google's hard drive survey, Microsoft's hard drive survey, and others, we can reasonably assume that the margin of error on the Backblaze data set is probably at least 5C; well above our less than 0.5C variation.

Their "takeaway" is that hard drive temperature has zero correlation with failure rate. However, that flies in the face of major scientific studies on motors, bearings, electronics, hard drives, etc, which actually show temperature has a causal relationship with failure. I don't particularly want to turn this thread into a Backblaze discussion, so I won't say any more about it here, but I implore you to be critical about data. Just because the authors throw around p-values does not make it statistically sound.

 

[Robert Trevellyan]

I would say that the idle timer is significantly more important than TLER

These forums have seen cases where a failing drive makes the entire FreeNAS unresponsive, not just the storage. This can make identifying and replacing the failing drive quite challenging.

When I amortize the cost difference between Greens and Reds over their expected life, not forgetting the extra year's warranty, the Reds are an easy win. YMMV.

 

[Bidule0hm]

I'm in this case right now... I've a failing drive (it's maybe just a bad contact on a connector because the drive isn't too old and the last SMART report didn't show anything wrong, I didn't check what's the problem for now because I need to correct one of my script before, well, enough of my life :p) and the GUI answer with a not found page error... But SSH and datasets works perfectly so not a big deal.

It's weird because I use only NAS drives so they should have TLER, seems I'll need to check that to be sure.

Edit: well, the drive is dead. It even prevented the server from booting. Tried to plug it to a desktop to see if I can extract some SMART values --> hard drive error on POST.

TLER is ok (7 seconds) but as the drive failed very badly I guess TLER wasn't useful in this case. I've seen in the log that Python had segfaulted less than 10 min after the drive died, that's why the GUI wasn't responding at all...

Some data for future reference: 10450 hours (only one year and two months of usage...), 28 °C, no reallocated, uncorrectable or pending sector, no ECC error, LCC: 16683 (SMART data from 10 hours before the drive died).

 

[Woodward]

Well, I went with the Greens. It was price, since I saved almost $200 on six drives. I'll look for the utilities for TLER support and idle parking. Thanks for the advice. Now I need to stop getting so much work so I can put this thing together! Spending every day driving hundreds of miles fixing other peoples' computers (and scanner/scales, and laser printers, and pin pads, and receipt printers, and monitors, and phone systems) and the cobbler's children go barefoot.