garycase

Sure -- it's just a cage to hold the drives; it's the controller you use with it that determines whether or not you have support for drives > 2TB

"... when we update the wiki ..." ==> :) Does this imply that you intend to start keeping the wiki current !!?? If you go to the Wiki today you'll find the "Official Manual" along with a note that it is "... version 4.7 only " :) I agree ... but I also think that given a choice many would prefer a PDF manual that they can peruse at their leisure, whether or not they actually print it (most would NOT do that).

We are not going to maintain our docs in both wiki and PDF form at this time. Franks efforts here are much appreciated, but PDF versions of our docs will need to be community maintained for now. I wasn't suggested you maintain the PDF => only that you host it. I don't see the point. Whether it's hosted by us or not really doesn't matter. Just the convenience of having a direct download on the site ... I agree it doesn't really matter [the same thing could be accomplished by simply having a link to the Dropbox account where it's hosted now]. A casual user who isn't active in the forum wouldn't even know the PDF existed if he/she clicked on "Support - Documentation" on the LimeTech web page.

We are not going to maintain our docs in both wiki and PDF form at this time. Franks efforts here are much appreciated, but PDF versions of our docs will need to be community maintained for now. I wasn't suggested you maintain the PDF => only that you host it.

Thinking about this a bit -- Limetech should host this PDF !! I can't think of any reason why they wouldn't want to have an easily downloadable manual available for all of their customers. The only issue is the currency of it -- as the online manual changes, the PDF would be out-of-date ... but an "as of" date for both the on-line manual and the available PDF would help mitigate that => and may encourage someone [either at LT or a volunteer like Frank ] to update the PDF from time-to-time.

Nicely done Frank. I'm also an old codger who likes to have the info immediately at hand with having to find the link. I simply put the PDF on my 2nd monitor so it's easily at hand instead of printing it, but the concept is the same.

I suppose anything's possible, but I've never seen it on ANY disk ... Seagate, WD, Hitachi, Maxtor, Toshiba, Samsung, or any other drive I've ever had. And note that the parameter is called "Power On Hours". In addition, if it WAS in minutes, then it's showing less than 54 days of power-on time [77170 / 60 => 1286.17 hours ... 1286.17 /24 = 53.59 days] I seriously doubt that's what it's showing. I think you simply have an erroneous SMART counter.

That does indeed show 77,10 hours = almost 9 years. This cannot, however, be correct for a 2TB drive, as they simply haven't been available that long. Here's a 26 Jan 2009 article noting the arrival of the first 2TB drives: http://www.engadget.com/2009/01/26/western-digitals-2tb-caviar-green-hdd-on-sale-in-australia/ So either the SMART data is incorrect or you're looking at a different drive than you think. The largest drives that could be 9 years old would be 750GB drives. The value of 12 is probably correct -- this decreases from 100 as the drive ages ... and 77170 hours is certainly a very old drive.

Are you looking at the power-on hours value? If so, that value would indeed be close to 9 years -- but it can't be correct for a 2TB drive, as drives that size weren't available 9 years ago. FWIW 9 years ago the earliest 750GB drives were just coming on the market -- I bought a couple for ~ $250 each. 1TB drives were about a year later, with 1.5TB and then 2TB drives following about a year apart. So the oldest 2TB drives today would be about 6-7 years old. If you have a 2TB showing that many hours something is wrong with the SMART data. Trust me, it wasn't all that much. Modern drives (anything in the past decade) are VERY inexpensive historically. The first hard drive I bought for my home was a 26MB (Yes, MB) Seagate that, after a 10% discount, cost me $4500. That's about $173 MILLION per TB ... and that was in 1980, so in today's dollars that would be over half a billion dollars !! [$530 million to be precise] A 2TB drive at that rate would cost over a billion dollars

Note that in the setup you describe in that post you are NOT running with "... no fans at all ..." => you've blocked all intake paths except the front, and have a fan pulling air through the drives from the front. While this may indeed work a bit better than using the side fans with the leakage that the base design allows around the drive cages, it's no better than simply fixing that leakage with a baffle like Daniel did => and then you don't need to keep the front door open (or remove it). One of the attractions of the DS380 is the really clean look with the front door closed ... if you don't mind losing that, then what you've done works fine => but it's decidedly NOT "no fans at all."

Agree ... I was going to comment on that as well. The only real issue with the airflow is that a lot of air can bypass the drive cage => and the very simple skirt you added (which many others have copied) completely eliminates that issue, and keeps the drive temps very nicely controlled.

My PC-Q25B sits on a shelf about 5 ft away from my ears when I'm sitting at my desk -- and I can NOT hear it at all. If fact, I can't tell if it's on or not without looking at it. I suppose it may depend on the noise level of the drive you've got installed; and there may be some variance in the fans they provide, but it's still what I consider the best of the mini-ITX case choices. In any event, it's not at all difficult to swap the fan if you should decide you need something quieter. It's very easy to build in as long as you use an SFX power supply.

Note that spin-up groups have absolutely nothing to do with user shares => whether you use them or not has no bearing on the problem the spin-up groups resolve. I also keep all files associated with a given video on the same disk. But if my wife is watching a video on the main TV, and we have a guest (son, grandson, friend, etc.) who decides to watch a video in another room, if the video they want to watch is on a different disk, then I don't want the action of spinning up that disk to cause a freeze on the video already being watched => but this will happen in many cases if they're on the same controller. My drives are all assigned to spin-up groups that I carefully identified (as I outlined above) ... so this never happens. Considering there are only two of us, it's clearly not a big deal => but it's not all that unusual for my wife to watch a movie while I'm "fiddling" with the server (adding content) and occasionally we'll have a couple grandkids who are watching things in other rooms as well. So it's very handy to have the spin-up groups.

No need to change the default => drives > 2TB are automatically 4k aligned.

A word on the various comments suggesting changing this to apply to shares instead of specified disk groups => DON'T !! Spin-up groups are very useful if you have controllers that "freeze" activities on attached drives while another drive on that controller is being spun up (this is not uncommon). But if you have, say 10 drives in a "Movies" share, you certainly don't want all 10 of those drives to spin up when you access the share. On the other hand, if a media item you're streaming is on disk x, and you know that any access to disk y will cause disk x to "freeze" until y spins up, you'd most likely want to have x and y in a spin-up group, so when x spun up, so did y ... thus eliminating any hiccups in the streaming. It's not a critical feature -- this is true of a LOT of the "bells and whistles" that are in the current release -- but it's VERY handy for those of us who use it. It's very easy to test whether it's useful for your particular setup => spin down all drives; start a stream from one drive; and then, one-at-a-time, do an access to other drives and see if any of those accesses cause the streaming video to pause. If so, you've found an interaction you can eliminate with a spin-up group. Clearly it takes a while to identify all of the interactions -- but once they're all "mapped" you'll never again have a movie freeze because somebody else accessed the server.

That's correct. As long as you don't create the top-level directory UnRAID won't either if the share is limited to one disk. You should be fine. One caveat: r.e. "... if I were to remove a disk from a user share ..." ==> If you simply change the share settings to remove the disk from the share, that does NOT actually delete the information ... i.e. the top-level folder and all of its data would still be on the disk. So if you then create a new share with the same name on a different disk, THAT would result in the condition that was described earlier. But other than that, if you do as you've outlined, you'll be fine.

As noted above, you can set which disk(s) a share will use by either setting the "Included" disks or the "Excluded" disks. You don't need to use both of those -- just do whichever is most appropriate for your share. e.g. If you want a particular share to use all of your disks EXCEPT for a couple, just set those as "Excluded". That way if you ever add another disk to your array, it will automatically be included in your share. ... but if you want a disk to ONLY use a specific disk (or disks), then just list those in the "Included" disks -- no other disks will then be included in the share.

Nice build. The DS380 is indeed a very nice case and, as you know, blocking the air "leakage" path for the drive cooling fans is simple to do => it's unfortunate that it's necessary, but several folks on this forum have done the same thing with anything from cardboard to custom-fit plastic parts. A shame Silverstone didn't just design it right

That's FAR more power than you need. 650 watts would be plenty.

The TDP for your CPU is FAR more than the CPU will typically draw => that's the max power when the CPU is running at 100% load, which, with that CPU, will be VERY rare (if ever) in UnRAID. The chipset has a TDP of 7 watts, and the rest of the motherboard's draw will be pretty nominal. Then you'll have the memory modules, any add-in cards, and your GPU. I think any power supply that has enough current capability to handle the startup load for 24 drives is going to have PLENTY of total power, as the actual consumption will be far lower than that initial load. The biggest potential "extra" load you'll add is if you decide to install either a high-end graphics card; or perhaps multiple graphics cards (for VM pass-thru). I'd think a quality 750w unit is plenty for this system, but you could go to 850 if you want to have even more "headroom". The Corsair HX or AX series, or a Seasonic X series unit would all be good choices. You absolutely want to use a quality UPS unit ... this almost completely eliminates the likelihood of the kind of failure you're concerned about, as MOST power supply failures are either caused by power spikes (virtually eliminated by the UPS) or by startup transients (during which time the drives are all just spinning up ... no writes are in process).

The full SMART report would show more; but there's enough info in what you posted to say that the drive is clearly bad and should be tossed. The very large # of pending sectors indicates that there are a lot of "uncertain" sectors that aren't reliably writable. The fact this number changed so much indicates that when zeroes were written to them, the writes were successful; but then 9 of these had read issues in the post-read so you had 9 pendings at the end of the pre-clear. You've only got one re-allocated sector; but with the large number of "flaky" sectors that are being marked as pending, and then cleared on the next write; and then sometimes being re-marked, etc. the drive clearly is NOT reliable. In addition, it's got a "Failing Now" parameter => if you have ANY SMART parameters marked with this you should trash the drive. In other words ... Toss it

If you're certain that's 100% reliable, then that will certainly do the trick as well. I'd do it enough times in a row to be confident about it, however. i.e. wait for auto-sleep; wait a few minutes; then press the power button. Repeat 3-4 times. Then repeat the whole thing manually putting it to sleep state via the GUI button. If it is definitely working perfectly this way, then that's a reasonable solution. If that's the case, I'd disable WOL in the BIOS ... this will eliminate any chance of accidentally causing the problem.

In that case, I'd just disable S3 sleep. The system will simply spin down all the drives, but otherwise still be on, so you'll not have any issues. Yes, it'll use a bit more power, but it'll be 100% reliable. Meanwhile, you can continue to look into why S3 isn't working correctly, and if you find a solution, you can set it up again -- but assuming reliable operation is more important, I'd just disable it for now.

What's the difference in what you did with #3 vs. #2 ?? If it simply didn't wake when you did the WOL in #3, try sending the magic packet again instead of using the power switch. Sometimes sending the packet 2 or 3 times in a row will be more reliable than depending on a single instance. #4 is likely the most reliable method => it guarantees the system has been properly shut down; and you're simply using WOL to turn it on. No S3 sleep state involved.

If the 2nd sequence doesn't work (i.e. you can't wake it via WOL), then clearly the system is NOT going to S3 state properly. This explains why UnRAID is doing a parity check ... the system is simply losing power and not shutting down properly. What happens if you do the following: (a) Use the Web GUI to Stop the array and then Power Down the system. (b) Try to turn it on via WOL