|
Is shucking hard drives a good strategy? I am building an Unraid server, but only have 6 bays, so I don't want to use 2 or 4 TB drives. Can I get some advice, goon sirs?
|
#
?
Sep 14, 2018 00:03
|
|
|
|
| # ? Apr 26, 2024 06:20 |
|
|
Yes, shucking is a good way to get solid drives at lower prices. Just be careful not to break the enclosures and keep them around until the warranty is up. WD Easystore 8TB externals regularly go on sale for ~$150 or less and are essentially WD Red NAS drives inside. Be careful with the Seagate externals, especially the "archive" line, since they are SMR (shingled magnetic recording) drives and not really suitable for RAID applications.
|
|
#
?
Sep 14, 2018 00:09
|
|
|
Hitachi NAS drives for performance and probably the best RMA rates.
|
|
#
?
Sep 14, 2018 00:11
|
|
|
Fancy_Lad posted:I'm finishing up a migration of ~30TB of data from a Windows+DrivePool installation to UnRaid. Also, how crucial is a cache drive if I'm only looking at writing ~5GB per month to it? From everything I've said, that doesn't seem mission critical either. Lastly, what's the best way to setup unRAID to be accessed outside of your LAN. This is basically an offsite backup for my office, plus a few personal things in its own container. 
|
|
#
?
Sep 14, 2018 01:29
|
|
|
eightysixed posted:Question on this, and I've read opinions that oppose eachother. I should have all my new hardware on Saturday for an unRAID box, and am only moving over about 4TB worth of data. I've read to set up unRAID without the parity drive, copy all the files over, and then add the parity drive to get things copied over quickly. I don't really care about the speed of the transfer of the data, it's only backups. If I setup unRAID with the parity drive first,how long do you think we're talking about moving a mere 4TB? I can handle a day or two, and there's no way it should be longer than that. 1. You're going to have to run preclear on all the drives so make sure whether you add the cache or not to start that you preclear it. It's going to take a while so strap in. If you have the hard drives now and a spare machine, do the preclear now. If you add parity later, make sure you run a parity check immediately afterwards. 2. You don't need a cache drive in that instance 3. BitTorrent Sync
|
|
#
?
Sep 14, 2018 01:34
|
|
|
eightysixed posted:Question on this, and I've read opinions that oppose eachother. I should have all my new hardware on Saturday for an unRAID box, and am only moving over about 4TB worth of data. I've read to set up unRAID without the parity drive, copy all the files over, and then add the parity drive to get things copied over quickly. I don't really care about the speed of the transfer of the data, it's only backups. If I setup unRAID with the parity drive first,how long do you think we're talking about moving a mere 4TB? I can handle a day or two, and there's no way it should be longer than that. Also following up with my fresh experience 1. I did all my data migration with parity enabled, but when I was shuffling data to encrypt the array I scrapped the parity until that process was finished and immediately rebuilt the parity when it wad done. There's two ways to configure how UnRaid writes parity. The default read/modify/write was about 1/3 the speed of reconstruct write for me, although understand the downside of keeping all your drives spinning (I do this anyway, so no downside for me). With reconstruct write, I get roughly 130MB/s write speed if the array is doing nothing else. Here's the kicker with UnRaid, if your array is doing any other writes, even to another drive, all writes are going to slow down because your parity has to write the changes and will bottleneck you. Also in reconstruct write mode heavy reads will impact your write speed as well since it is reading all drives in the array. The file shuffle for the encryption just tanked my speeds so instead of letting that take 4 days I had backups so I ran with it. All my copies were from a local VM with passthrough disks and 4TB with nothing else going on would ballpark half a day. Matt Zerella is correct that you will want to preclear your drives and that will take you probably that long or longer (depends on the size). 2) Agreed on the cache drive. I have 2 SSD cache drives but I'm using them just for docker/VM storage and not write cache to any share... From what I've read off the forums this isn't unusual. The ZFS folks will probably be laughing at those transfer numbers above, but for a backup or media server you generally aren't going to care about speeds much. 3) I haven't figured this out yet. My desire is less remote file transfers and more just remote management... I have a "todo" to try to setup the OpenVPN docker, but I expect that won't be available if the array is stopped so that isn't really ideal there. Would also like suggestions that aren't "open your web interface or SSH port to the interwebs" 
|
|
#
?
Sep 14, 2018 05:18
|
|
|
8-bit Miniboss posted:Anyone have recommendations for 8 bay with hotswap, mini-itx cases? I got a hold of a Supermicro X10SDV-F by way of a goon coworker that I'm looking to transfer my Unraid setup to from my Lenovo TS440 as the HDD expansion kit is stupid rare to find and no one is selling theirs without asking for an arm and a leg. I have two ds380s, and I think they're pretty great. There are some mods you can do to the drive cage to alleviate heat issues, but I did nothing to it, as I use them as backup boxes that are rarely powered up.
|
|
#
?
Sep 14, 2018 09:26
|
|
|
I bought an 8 TB EasyStore because it was on sale, and I figured I might as well test the drive while it's still in the enclosure. I tried running the RD-Read test with HDD-SCAN, but it seemed like after a while it just started reading all the blocks as bad. The first time it was after 4%, the next time it was after 16%. I'm guessing that not all the blocks are actually bad. Is there some other tool/test I should be doing? Or is it possible something else is happening, like my machine is trying to spin down the drive after a set period and that messes with it? HDDScan has stopped at 16% and stopped incrementing the LBA and KB/s is now at 0, but it continues to add more bad 'B' blocks to the map. 
|
|
#
?
Sep 18, 2018 02:50
|
|
|
cheque_some posted:I bought an 8 TB EasyStore because it was on sale, and I figured I might as well test the drive while it's still in the enclosure. I tried running the RD-Read test with HDD-SCAN, but it seemed like after a while it just started reading all the blocks as bad. The first time it was after 4%, the next time it was after 16%. I'm guessing that not all the blocks are actually bad. Is there some other tool/test I should be doing? Or is it possible something else is happening, like my machine is trying to spin down the drive after a set period and that messes with it? Did you try WD's own tools?
|
|
#
?
Sep 18, 2018 02:57
|
|
|
I've just used WDs tools on my 4 drives (before I shucked), they are still alive.
|
|
#
?
Sep 18, 2018 03:43
|
|
|
Got my Dell H200 flashed over to IT mode! After loving it up, (gotta be my true self and all) got it working! I need to get better about reading the loving instructions.
|
|
#
?
Sep 18, 2018 03:47
|
|
|
Ah, good idea. Hadn't thought of that.
|
|
#
?
Sep 18, 2018 03:52
|
|
|
cheque_some posted:I bought an 8 TB EasyStore because it was on sale, and I figured I might as well test the drive while it's still in the enclosure. I tried running the RD-Read test with HDD-SCAN, but it seemed like after a while it just started reading all the blocks as bad. The easystores have a middleware board that sits between the disk and your USB port that filters commands. Despite it showing up as mass storage there is not a direct connection. Once shucked you will see what I mean. It prevents you from doing things like partitioning the disk.
|
|
#
?
Sep 18, 2018 04:08
|
|
|
Moey posted:I've just used WDs tools on my 4 drives (before I shucked), they are still alive. Same here, I used WD's tools to check the drives and used CrystalDiskInfo to check the model of the drive inside before I shucked all of them. Luckily all 8 of mine were fine and despite being bought at different times, batches, stores, prices, etc they all ended up being either the White 256MB EMAZ or the Red 256MB EFAX drives. Figured it would be a lot easier to do it that way vs trying to put the things back together and return them if I had gotten the less desirable drives.
|
|
#
?
Sep 18, 2018 05:37
|
|
|
is it possible to buy the ribbon cable that feeds the front panel on a Diskstation DS1815+? pretty sure I have accidentally crimped it beyond functioning
|
|
#
?
Sep 18, 2018 09:40
|
|
|
priznat posted:Case I have looked at for NAS stuff is the Fractal Designs 804: https://m.newegg.com/products/N82E16811352047?Keyword=Fractal%20design%20804 Eletriarnation posted:It's hard to beat the Node 804's ten 3.5" bays at the same cost or size, let alone both. I'm using an old full tower case that I don't want to throw away or use anywhere else, but I bought a microATX board for it partly so I'd have the option to use an 804 if I ever want to add more drives than would fit right now. redeyes posted:That and in no way is it easy to swap drives. Uglay. Oh my God, I regret this decision. Yeah, it's pretty terrible, you're right  At least they're all brand new drives so I shouldn't have to worry about it for awhile (hopefully), and it's maxed out.
eightysixed fucked around with this message at 02:41 on Sep 19, 2018 |
|
#
?
Sep 18, 2018 14:23
|
|
|
I’ve got at least 24 drives at home that have been humming along and I don’t really regret shoving 10 drives into my Lian Li A04 case. Now, I also have a U-NAS case that has hot swap bays, but I’ve needed to rotate drives out of there a grand total of... once, which was when I moved them from 2 TB drives to 8 TB drives. My normal maintenance repair replacements have been all in my Lian Li mini tower. I’ve gotten a habit of buying ITX boards for a while now but given the hardware hassles related to the form factor (nowhere near as much selection of server boards) I’d rather stick with microATX in the future. The Node 804 is a pretty decent case for non-NAS duties too.
|
|
#
?
Sep 18, 2018 15:59
|
|
|
eightysixed posted:I bought this, thanks, thanks the for the recommendation Yeah like you said, hopefully you wont have issues with failed drives for a long time. Mark the drives compared to what your OS sees so you can easily find a drive that might have failed.
|
|
#
?
Sep 18, 2018 16:53
|
|
|
Good idea. I actually do like the exterior looks of the case, but the HDD management is terrible.
|
|
#
?
Sep 18, 2018 21:33
|
|
|
 Does it have bad sectors or not, make up your mind you stupid computer.
|
|
#
?
Sep 19, 2018 00:10
|
|
|
Fancy_Lad posted:Matt Zerella is correct that you will want to preclear your drives and that will take you probably that long or longer (depends on the size). Matt Zerella posted:1. You're going to have to run preclear on all the drives so make sure whether you add the cache or not to start that you preclear it. It's going to take a while so strap in. If you have the hard drives now and a spare machine, do the preclear now. If you add parity later, make sure you run a parity check immediately afterwards. This appears to be a deprecated thing, as stated by SpaceInvader One himself as well as a thoroughly documented thread on the unRAID forums as of unRAID 6.4. Thoughts? edit: Oh, it's the plugin. So work from terminal? I have four drives in now, none assigned to an array, a brand new build up and running. I apologize for being a newbie with unRAID, this is just my first rodeo with it. Plow on, or do some sort of CLI precleaning? If so, what/how? 
eightysixed fucked around with this message at 02:36 on Sep 19, 2018 |
|
#
?
Sep 19, 2018 02:07
|
|
|
If preclear is depreciated then by all means just add the drives to an array and go.
|
|
#
?
Sep 19, 2018 02:27
|
|
|
I tend to agree, I just wanted to double check.
|
|
#
?
Sep 19, 2018 02:40
|
|
|
H110Hawk posted:
I ran into the same thing yesterday while it checked consistency/repaired after replacing a failing drive. It only had one phantom bad sector+UNC error though.
|
|
#
?
Sep 19, 2018 03:14
|
|
|
I have 2 WD Reds in a ZFSonLinux mirror (CentOS) and a spare SanDisk 64GB SSD sitting around doing nothing. The main times I actually see the Reds in action is during Plex streaming. The rest of the time the server is doing scheduled backups etc. Would installing my spare SSD as an L2ARC give an appreciable performance gain or is it not worth the effort? Most of the time I'm streaming 1080. Very occasionally I might stream a 4k movie. Would the SSD make a difference when you leave the room, then come back and rewind a Plex stream a couple of minutes? The system is pretty quick anyway and it's certainly not annoying waiting a second or two for the stream to catch up, but pimping it up would be cool if I could see a difference.
|
|
#
?
Sep 19, 2018 23:03
|
|
|
apropos man posted:Would installing my spare SSD as an L2ARC give an appreciable performance gain or is it not worth the effort? Most of the time I'm streaming 1080. Very occasionally I might stream a 4k movie. Short answer: no Long answer: it depends on your workload, but due to the way ZFS's cache tiering works you generally want to be maxed out on RAM before you even think about L2ARC. You want a minimum of 64 GB of RAM before you do L2ARC and your L2ARC should generally not exceed 5x the size of your ARC (depending on block size). apropos man posted:Would the SSD make a difference when you leave the room, then come back and rewind a Plex stream a couple of minutes? The system is pretty quick anyway and it's certainly not annoying waiting a second or two for the stream to catch up, but pimping it up would be cool if I could see a difference. Lol definitely no. I assume the problem is that Plex is destroying your transcoded file as it plays, and when you rewind it has to go and rebuild the buffer. Putting the file on an SSD before you play it would probably help, but I dunno if Plex can do that. You can also use vmtouch to pin a file into memory, which would do the same kind of thing.
|
|
#
?
Sep 19, 2018 23:11
|
|
|
Ah. OK then. I'm only running 16GB RAM as it's a home server so I won't bother going through the rigmarole of shutting everything down, getting the server out of the cabinet and messing around with drives. I think part of the idea was a bit of posturing just to think "Yeah, I'm using an L2ARC!", Lol.
|
|
#
?
Sep 19, 2018 23:18
|
|
|
apropos man posted:Ah. OK then. I'm only running 16GB RAM as it's a home server so I won't bother going through the rigmarole of shutting everything down, getting the server out of the cabinet and messing around with drives. I thought the same thing at one point, it was like "hey I've got an extra SSD laying around, can I maybe make use of this?" and the answer is no. ZFS doesn't really work like that and L2ARC has its downsides too.
|
|
#
?
Sep 19, 2018 23:22
|
|
|
I also just cut down on the energy use by... putting a Xeon in the server! I'd been running an i3 6100 in it for a year or so, which from memory is a 54W TDP chip with no turbo boost, so it's permanently pegged at 3.7 GHz. I used that particular CPU because it allowed for ECC RAM when I originally built the server. As I've gradually got it running more tasks/VM's I was wondering about running extra cores in there and noticed the Xeon E3 1240L v5, which has 4c8t, much more cache, turbo boost to allow it to idle better and only a 25W TDP. It's more or less an energy efficient i7 without iGPU. I the maximum clock speed is lower than the i3, so single-threaded performance is slightly lower but overall I'm pretty pleased with it. (Numbers may be slightly off, as I'm posting from my phone and can't be bothered to look them up)
|
|
#
?
Sep 20, 2018 00:00
|
|
|
If you have a WD MyCloud on the internet you're probably already owned... https://www.theregister.co.uk/2018/09/18/remote_access_vulnerability_western_digital_my_cloud/
|
|
#
?
Sep 20, 2018 02:21
|
|
|
apropos man posted:I also just cut down on the energy use by... putting a Xeon in the server! The lack of turbo on a cpu doesn't stop it from throttling down to lower power consumption. I think Intel's buzzword for it is SpeedStep, and the i3-6100 has it (also idle states). That said, Xeons are neat and the low power ones are pretty amazing.
|
|
#
?
Sep 20, 2018 03:03
|
|
|
Oh. Is that something akin to the CPU being rapidity triggered between different clockspeeds? I hadn't realised all this time that it was a non-turbo. As far as the Xeon I bought goes, there's another model: the E3 1260L v5, which clocks a fair bit higher than the 1240L that I bought. I was thinking the other night that I should have got the faster clocking one. Then again, it's TDP would have been almost the same as my i3 and my server's really not doing much the majority of the time. From a geek perspective I should have gone with the faster one but in terms of practicality the one I'm using is more than enough. I bought it from an eBay seller in China and it arrived at my door in the UK in six days, which surprised me.
|
|
#
?
Sep 20, 2018 03:14
|
|
|
apropos man posted:Oh. Is that something akin to the CPU being rapidity triggered between different clockspeeds? Besides clock speed, the extra cache on a Xeon is often a speed improvement . It depends on what tasks the CPU is doing but I've seen folks claim that every MB of cache is like 100mhz of clock speed. Those people really liked their Xeons, though, and I think that that number would not hold up in a lot of scenarios.
|
|
#
?
Sep 20, 2018 03:42
|
|
|
The main benefit of Xeons is ECC, they're otherwise equivalent to Core chips (and there are -S and -T variants of the Core chips that reduce TDP as well, but neither are binned any better than the standard desktop chips, they just have a lower power limit permanently burned in, you can drop the power limit on a normal chip and get the same result). I think most Xeons that have Core counterparts have the exact same amount of cache as their Core counterparts but I won't say that is 100%. The ones that really go nuts on the cache are the E5s, which again have counterparts in the HEDT lineup with the same amount of cache. eg the E5-1650v3 is the same thing as a 5820K only 100 MHz slower, and both have 15 MB of cache. The E5 Xeons also have the benefit of RDIMM support for truly excessive amounts of RAM, which isn't on the E3s since they are derived from desktop chips that don't need that. Anyway, 1 MB of cache here or there certainly isn't going to make up for the unlocked cores being able to push 1.4 GHz higher or whatever. But, it certainly is nice for gaming, heavily-OC'd X99 chips still top the charts in many titles because they have way way more cache than the client chips (and unlike Ryzen they are a monolithic cache vs the "2x8 MB" style configurations, which introduces some thrashing issues in some cases). Shame Skylake-X moved to the mesh interface because they perform like rear end in gaming, X99 is really the end of an era. Paul MaudDib fucked around with this message at 14:43 on Sep 20, 2018 |
|
#
?
Sep 20, 2018 14:39
|
|
|
For anyone who doesn't want to shuck, Newegg has 8TB Toshiba N300 drives for $360 per pair of two: https://promotions.newegg.com/NEemail/Sep-0-2018/HotDeals-Xklwiq21-20/index-landing.html
|
|
#
?
Sep 20, 2018 16:57
|
|
|
IOwnCalculus posted:For anyone who doesn't want to shuck, Newegg has 8TB Toshiba N300 drives for $360 per pair of two: https://promotions.newegg.com/NEemail/Sep-0-2018/HotDeals-Xklwiq21-20/index-landing.html I was just considering grabbing a pair. Is there any RMA data for these drives?
|
|
#
?
Sep 20, 2018 17:12
|
|
|
redeyes posted:I was just considering grabbing a pair. Is there any RMA data for these drives? Backblaze thinks Toshibas are above-average reliability. They're tracking desktop drives but if anything NAS drives should be better. I've got five of the 6 TB X300s (they were the best value before the era of shucking Reds) and they've all been fine even under some rough usage. They are not quiet, and Bryan Cantrill dug up some potential firmware issues (that can affect write speeds, not corruption), but overall they are decent drives. (again, personally haven't seen any problems, they'll happily do 130-150 MB/s for me, so  )
Paul MaudDib fucked around with this message at 18:07 on Sep 20, 2018 |
|
#
?
Sep 20, 2018 18:02
|
|
|
I haven't heard much of them one way or another. I do have four of the 5TB X300 drives with zero faults of any sort so far.
|
|
#
?
Sep 20, 2018 18:02
|
|
|
Cool. So not Seagate. lol
|
|
#
?
Sep 20, 2018 18:43
|
|
|
|
| # ? Apr 26, 2024 06:20 |
|
|
I’ve got four of the 4 TB Toshiba desktop drives next to four WD 4 TB drives in a RAIDZ2 and they’ve been perfectly fine. They do run a little warmer and aren’t the best for noise comparably but my bigger concern is reliability than anything cosmetic / aesthetic when you’re talking 8+ drive arrays.
|
|
#
?
Sep 20, 2018 18:45
|
|