What drives are you using?

[jgreco]

5TB WDC RED's are the sweet spot at the moment and have worked well for me. I've had more issues out of the 4TB Seagate NAS drives. Just my $0.02.

We've had ....so-so... luck with the Seagate desktop 4's but the NAS 4TB's have been good. Also, while the 2TB Greens are a frickin' fail-mare, we've had great luck with the WD Reds in varying capacities including the 1TB 2.5" Red.

From a price point perspective, I'd bet the 5TB drives are where it is at. Do remember when you're looking at drive costs to factor in the cost of parity and the cost of the system itself, because you may come to very different conclusions than if you just look at raw drive costs.

 

[craigdt]

Sure the 6TB are a little expensive, but they will most likely get cheaper in the future as larger sizes come out as well, personally I wouldn't touch the archive drives except maybe for the backup NAS but even then I'm not sure if I want to be the guinea pig ;)

 

[jgreco]

Mostly a matter of what you're going to use it for. I'm pretty sure the SMR drives are the future of spinny rust, because flash just can't touch the price of platters, but fast access to data is rapidly moving to flash.

 

[HoneyBadger]

I cannot justify it right now as the big pool here is only at 48% capacity ...

"That's awful close to the 50% recommended upper bound for optimal performance, don't you want your pool to be running at maximum efficiency?"

There. Justified. Warm up the credit card. ;)

 

[jgreco]

"That's awful close to the 50% recommended upper bound for optimal performance, don't you want your pool to be running at maximum efficiency?"

There. Justified. Warm up the credit card. ;)

Yeah, except that I've got to justify it to the boss, and that awful bastard actually knows ZFS. He's not likely to sign a PO for it.

 

[AgileLogic]

I've been using WD Reds -- WD40EFRX and WD60EFRX, and they've been good to me over the past 3 years.

But, my usage is nowhere near intense, and I don't have very many drives in use even over the past 5 years. Over the past 2 years, I had a half dozen 2 GB Seagates fail in succession, so I stay far away from Seagate now.

Backblaze has been tracking lots of stats on their drive reliability in their big data center (~50,000 drives). You can find their latest report here: https://www.backblaze.com/blog/hard-drive-reliability-q3-2015/

Their data shows the HGSTs are the most reliable, sometimes by a large % over other brands. Followed by WD. They also show a much larger failure rate on Seagates (an order of magnitude higher in some cases, 10% vs 1%), matching my experiences. Interestingly, they use an awful lot of the cheaper NAS class drives, not the enterprise class drives. Also interesting is 4 TB drives show a lesser failure rate than 6 TB drives across multiple brands, even for those where Backblaze has a similar time in service for both.

I'm about to embark on a new FreeNAS build with 16-24 drives, and I'm leaning towards the HGST Deskstar NAS 4 TB. They seem to have a decent price/capacity point (the WD Reds are a bit less expensive), and if the Backblaze data holds they might be a good price/reliability point as well. If I want more capacity for the number of drives, I might try the 6 TB HGST Deskstar NAS. The HGST 8 TB helium & 10 TB SMR drives seem too cutting edge for me right now, and are pretty expensive.

 

[Robert Trevellyan]

They also show a much larger failure rate on Seagates (an order of magnitude higher in some cases, 10% vs 1%)

This seems like a bit of a misread to me. Their data show high failure rates with one particular Seagate model, and they state that other models work well for them and that they continue to buy Seagate.

 

[jgreco]

I had more than half the WD 2TB drives we bought fail. Regardless, I still buy both WD and Seagate drives. All the little fsckers fail sooner or later. Certain batches are more craptacular than others.

 

[reqlez]

I had more than half the WD 2TB drives we bought fail. Regardless, I still buy both WD and Seagate drives. All the little fsckers fail sooner or later. Certain batches are more craptacular than others.

Hi. i'm getting really bad luck with WD drives as well, just started having issues recently... the older WD RED drives are fine. What i'm getting is drives are being dropped from array ( 3 different systems ) but there are no errors in logs at all and tests are okay. They just drop because they become "busy" ( timeout ).

 

[jgreco]

Are you running SMART tests on them? Drives shouldn't drop because they become too busy with legitimate traffic. That suggests maybe errors.

 

[reqlez]

Those drives have to be bad ... but, its intermittent. A drive drops every 3 months or so... and smart tests show nothing, no cable issues, no relocated...

 

[MtK]

Those drives have to be bad ... but, its intermittent. A drive drops every 3 months or so... and smart tests show nothing, no cable issues, no relocated...

Are these drives always in the same slot?
Maybe it's a cable/backplane/power issue...

 

[jgreco]

Those drives have to be bad ... but, its intermittent. A drive drops every 3 months or so... and smart tests show nothing, no cable issues, no relocated...

What's the specific cause reported when they drop?

 

[reqlez]

What's the specific cause reported when they drop?

Basically the drives ABORT commands eventually. Thats what's in the logs.

I tested this in another machine, outside of freenas ... and after about 3 or 4 passes the drive did ABRT a block read command.

This only seems and issue with newer WD RED drive manufactured in the last year or so ... i read that they made some new Firmware because some people wanted to see better "benchmark" results, for all I know it's and issue with FW but could be unstable electronics ?

Anyway ... i switched to HGST NAS drives, can't take a chance. ( by the way, I used Victoria HDD test windows application to do tests )

 

[jgreco]

Yeah, I'm sorry to say, if the drives are tossing ABORT's, that's a bad thing. It's possible there's some tunable that could affect that, but I just don't have time right now to go researching that. Might want to see if there's a way to limit the NCQ queue depth is what I'm *thinking*, but it's just a halfway-thought-out thought at this point.

 

[reqlez]

Yeah, I'm sorry to say, if the drives are tossing ABORT's, that's a bad thing. It's possible there's some tunable that could affect that, but I just don't have time right now to go researching that. Might want to see if there's a way to limit the NCQ queue depth is what I'm *thinking*, but it's just a halfway-thought-out thought at this point.

You are correct ... no drive should ever throw ABRT at a command ... unless the command is not supported by the drive for some reason, so I can't just continue using those WD RED Drives until I hear from somebody that the found the firmware issue or other issue, they will fix it, and will swear to stress test better next time.