4TB seagate external HDD $129 at Amazon

[bmacklin]

Much cheaper than any 4 TB internal drives out there right now. I am planning on getting the drive out of there and use it for my server.

Amazon.com: Seagate Backup Plus 4 TB USB 3.0 Desktop External Hard Drive STCA4000100: Computers & Accessories

supposedly uses ST4000DM000 drive.

Anyone have any experience with the ST4000DM000 with ZFS? Supposedly they spin down causing hardware raid problems, but wondering if they are fine for ZFS?

 

[bmacklin]

Much cheaper than any 4 TB internal drives out there right now. I am planning on getting the drive out of there and use it for my server.

Amazon.com: Seagate Backup Plus 4 TB USB 3.0 Desktop External Hard Drive STCA4000100: Computers & Accessories

supposedly uses ST4000DM000 drive.

Anyone have any experience with the ST4000DM000 with ZFS? Supposedly they spin down causing hardware raid problems, but wondering if they are fine for ZFS?

Expired. Should have bought 2.

 

[odditory]

Anyone have any experience with the ST4000DM000 with ZFS? Supposedly they spin down causing hardware raid problems, but wondering if they are fine for ZFS?

There are no issues with the Seagate STx000DM00x drives and a legit hardware RAID controller. I have many, many of them. They're fine with ZFS as well.

 

[33_viper_33]

Be careful with all newer drives with hardware raid. I have 3 of the original Seagate 4TB HDDs which work great in hardware raid. Sometime after I bought those I purchased 2 more to expand the array and have had issues ever since. The new drives have a different firmware which is too tolerant with sector errors. My areca raid card won't wait as long as the drive before determining there is a drive sector error before moving on. Ultimately the drive drops off line and the array rebuilds itself. This happens every 2 months or so which is why I'm now looking into software RAID. No issues in ZFS thus far, but I'm still in the experimental stages and haven’t had it up and running for long term stability yet. The older consumer drives have no issues. I have 9 x 1.5tb drives in RAID 6 that have been running for 4 years now. Recently, I've started seeing drive failures which is why I'm setting myself up to move to 4TB drives.

I believe this is part of the manufactures agenda to get people to purchase the more expensive enterprise drives. The only major difference between consumer and enterprise drives is the firmware and the tolerance of raid controllers in addition to longer warranties. Check out BackBlaze and their research. The company does online storage and backups using nothing but consumer drives to include these external models, pulled out of the stock enclosure and placed in some very cool rackmount storage chassis. You can start with this article http://www.servethehome.com/storage-reliability-figures-backblaze/.

 

[odditory]

Be careful with all newer drives with hardware raid. I have 3 of the original Seagate 4TB HDDs which work great in hardware raid. Sometime after I bought those I purchased 2 more to expand the array and have had issues ever since. The new drives have a different firmware which is too tolerant with sector errors. My areca raid card won't wait as long as the drive before determining there is a drive sector error before moving on. Ultimately the drive drops off line and the array rebuilds itself.

That's definitely odd, and is counter to my experience with 3TB and 4TB Seagates pulled out of external enclosures and Areca cards + HP SAS Expanders. I bought some as recently as last week to add space. If you want help troubleshooting this, create a thread with model of Areca and firmware rev, make/model/manufacture date/firmware of Seagate drives, HDD power management settings in Areca GUI, and other details like SAS expander make/model.

Also, before adding a newly purchased drive to a raidset for expansion, always at least run a full read/write of the surface with something like "badblocks -wvs /dev/sda" in terminal using PartedMagic -- otherwise you're exposing your existing raidset to unnecessary risk with an untested drive since if the expansion fails due to bad drive, the raidset is degraded state as you prob know.

I believe this is part of the manufactures agenda to get people to purchase the more expensive enterprise drives.

I dont believe they're that intentional or thought out about it. I realize how it might seem that way. I do know WD took steps to intentionally gimp the processor on their consumer class drives so they wouldn't be able to handle enterprise type workloads, but I dont think Seagate has done same. So yes and no that they're trying to cut consumer class drives short in some way, but I dont think dropout problem is actually an unrelated issue. Also, very little known fact: Areca issues APM disable command by default so consumer Seagate STx000DM001 will not load cycle (head park/unpark) excessively.

 

[33_viper_33]

This is a semi common problem and I've continued monitoring other sites for a solution. It has to do with TLER", "CCTL", and "ERC" pending the manufacture your talking about. Consumer drives try to recover bad sectors, which is good when talking about a single drive. With raid, the array has to wait for that single drive to do its thing. Many controllers aren't patient enough and will drop the drive. If you know how to reprogram the firmware or controller to get around this, I'm all ears (errr, eyes). I've been watching for quite awhile and have yet to see any solutions other than software raid where you can adjust the array's timeout value.

 

[odditory]

This is a semi common problem and I've continued monitoring other sites for a solution. It has to do with TLER", "CCTL", and "ERC" pending the manufacture your talking about. Consumer drives try to recover bad sectors, which is good when talking about a single drive. With raid, the array has to wait for that single drive to do its thing. Many controllers aren't patient enough and will drop the drive. If you know how to reprogram the firmware or controller to get around this, I'm all ears (errr, eyes). I've been watching for quite awhile and have yet to see any solutions other than software raid where you can adjust the array's timeout value.

Very well aware, could write a book on it. Unfortunately, on the net you'll find only misconceptions about what's referred to under the blanket term "TLER", and most people believing that one or multiple drives dropping from an array simultaneously must have something to do with drive stuck in a deep recovery cycle and not some other issue. ERC's are extremely rare.

Most of the time a drop is actually due to some other reason. There are many variables. I'd be happy to get into it further if you want to create a thread to troubleshoot any drive dropouts you may still be having. But I can pretty much guarantee its not because your drives are actually going into a deep recovery cycle, there's something else going on.

Example, with Areca cards there was a time that a false-positive timeout/drop could occur if staggered spinup was set too aggressively. Then there was the whole WDTLER debacle years ago where people would enable TLER on a green drive, and it had the placebo effect of reducing or eliminating dropouts, but yet the dropouts weren't caused by an actual deep recovery cycle and the TLER setting simply had the side effect of masking what was a WD firmware bug to begin with. A few examples, there are many.

 

[mrkrad]

You can increase the LSI chipset to 30 seconds but the problem is then your whole raid set (vm's sql server etc) will then have to deal with such a large timeout on disk i/o and they start to freak out.

I did have an idiot mount a tape drive on a RAID controller back in the SCSI days it caused a 2hour timeout, oddly windows recovered fine when the tape drive was powered off, but the apps like sql server were very angry for losing storage for so long!.

So there is many facets of the TLER/CERC that you have to consider! If you are running vm's on that drive set, then 8 seconds without storage is going to make the vm's themselves very unhappy!

End of the day, if you must run NEARLINE drives, stick to good quality SAS drives. RE4 SAS for instance, since it is multi-target (up to 16) you can ask the drive what it is doing when it is doing a recovery! with the sata version , you have only one target at a time, and the o/s cannot ask the drive what it is doing ! since it cannot accept multiple targets at a single time!

The cheap 4tb drives are good for JBOD or external DAS over usb/tb but otherwise i'd avoid them like the plague for mainline storage!