testdasi

No need to buy a sata card. Try this method first - passing through the scsi bus. 1. Put this in the xml right before the </devices> tag scsi_host9 in my case is where the optical drive connects. Change it to the appropriate value for you. If not sure, type lsscsi in the console command and it should be obvious. <controller type='scsi' index='1' model='virtio-scsi'/> <hostdev mode='subsystem' type='scsi'> <source> <adapter name='scsi_host9'/> <address type='scsi' bus='0' target='0' unit='0'/> </source> <readonly/> <address type='drive' controller='1' bus='0' target='0' unit='0'/> </hostdev> 2. Boot VM up and go to Devices Manager, you should see 2 new devices without driver 3. Install virtio scsi driver for the no-driver devices. You should only be able to install it for ONE of the 2. 4. Reboot and it should work. If it doesn't, double check that unRAID has automatically changed the xml codes correctly into the below. I have no idea why that happened, it just happened. <controller type='scsi' index='1' model='virtio-scsi'> <address type='pci' domain='0x0000' bus='0x00' slot='0x09' function='0x0'/> </controller> ... <hostdev mode='subsystem' type='scsi' managed='no'> <source> <adapter name='scsi_host9'/> <address bus='0' target='0' unit='0'/> </source> <readonly/> <address type='drive' controller='1' bus='0' target='0' unit='0'/> </hostdev>

I heard of a Roku 2 2015 edition which is faster than the original Roku 2. Is that true? What's wrong with the Fire Stick interface?

I have an Xbox One which can stream Plex easily. However, it takes 5 minutes from turn on to getting into the app - basically defeats the purpose ! So I'm consider to get an Amazon Fire TV or a Google Chromecast to receive the stream from my unRAID server. I also heard of Roku but not sure what's the diff. Key requirements are minimal cost + quick and easy. Compatibility is a obviously a must but given transcoding is done on the server by plex, I wouldn't imagine this to be a problem. So my questions are [*]What are your opinions about these devices [*]How much does it cost? Device price I can search on Amazon but the app / running cost is confusing. Plex officially says chromecast / Amazon TV support is free, but then the app reviews say it's not (or worse, doesn't work).

Rebuilding again won't work since parity was updated during the format. Yeah, that's what I suspected too so that's why told him not to do it until an expert confirms. Sounds like the 2TB is the only way to recover anything now. I wonder if there's anyway to add a red note to users before the click the "Format" button. This is the 3rd times I remember seeing a post about this and I haven't been an active member for long. Something like "Do NOT format if disk is still unmountable after a rebuild - ask the forum for further actions" would suffice I guess.

Long explanation - pardon my previous post in red. Hope I wasn't too late. Rebuilding an unmountable drive will result in another unmountable drive. Being unmountable suggests it is already corrupted (file system?). The rebuilding process simply re-create the drive as per what parity "remembers" and if parity remembers it as unmountable, it will return to that state. Parity check in itself does no harm. However, some might tick the "Correct any Parity-Check errors by writing the Parity disk with corrected parity." box. That basically tells the parity drive to re-remember your now-formatted disk as being empty. That results in zero chance of restoring any data as any parity info is lost. What I would do (DO NOT do it yet, wait for an expert to 2nd it, or give another suggestion - I'm just another user so I do make mistakes too). Rebuild the 4TB again Run xfs_repair as per johnnie.black's suggestion If you have already done a correcting parity check then I'm afraid your only recourse is to try to fix the original 2TB drive and recover as much data from it as possible.

DO NOT RUN PARITY CHECK!

Is this the sequence of event? 2TB drive fails and shows up as unmountable Replace 2TB drives with 4TB drives and rebuild 4TB drive still shows as unmountable after rebuild You format the 4TB drive 4TB is mounted but with empty share folder

The unRAID bridge is 10Mbps (>1GB/s) so you will not be bottle-necked by the connection (albeit with increased latency). If you want direct access to the drive (i.e. without any network overhead) then best way is to passthrough the SATA bus. You can passthrough just 1 bus (instead of the entire controller).

I really don't want to go into brand recommendation cuz the proverbial "YMMV" applies. My sample is also quite small, only about 10+ drives. Of which the only one that failed was a WD Black - but then it was because Amazon stupidly posted it in a thin cardboard envelope with zero padding. Personally, I have always put Hitachi as my first choice (it appears to be confirmed by Backblaze in 2015 (<1% failure rate vs WD 2.5% and Seagate 3%). But it is now owned by WD (rebranded as "HGST") so who knows where things are heading.

Not clear what you have done. I'm guessing someone probably will ask for logs + screenshot of your array config. Very first question: what do you mean by "data folders are still empty"?

Do they come from the same order? You might have a bad batch (or WD Red drives are crap, or you have bad luck, or any of those combinations ). I made it a point to never order the same drive from the same seller within a short period / same order. I read somewhere that drives on the same batch tend to have similar issues.

Question: Is there anyway to periodically mount a SMB share? I want to (automatically) mount an smb share from my workstation to my unRAID server for back up purpose. It would only mount every Friday night and unmount on Sunday night. That should be doable if there's a command line or something I guess.

I hope you didn't tick the automatic correction of parity error option. Formating the drive => you lose all the data on it. Correcting the parity at this point simply means the parity is updated to reflect the fact that the new drive is blank. That means you lose all your data on the drive, permanently. If this is indeed the unfortunate case, I recommend using a data recovery tool to recover the data on the failed 2TB (as much as you can). That's the only source of the data you have now. I think there needs to be a big notice somewhere to tell people something like "rebuilding a drive that is unmountable will still make it unmountable" - I didn't know about that and only read that on the forum, or at least that's what I remember reading.

Didn't work, it still shows up in Windows.

Let's say I have share called Documents with a subfolder called Contracts. Is there anyway that I can hide that Contracts folder and perhaps unhide it using the console? I'm trying to avoid having to have a separate share to hide / unhide the folder.

I am being over the top with RAM estimate. Dockers won't need as much RAM. But unlike windows, unRAID doesn't have built-in "swap file" functionality (as far as I know). So not enough memory => your running VM suddenly crashes without even a BSOD => it's almost like a physical computer gets a power cut. unRAID also fully utilises any free RAM as cache (not to be confused with the cache drive). So you are not really "wasting" much (e.g. I have 177MB free RAM right now i.e. 177MB of wasted RAM, not even half a percent). I built a server with the expectation of NOT upgrading in the next 5+ years (except for perhaps increasing storage space). 6+ years ago, I had 6GB RAM and my friends said it was an overkill. 6GB is barely just enough now. The Intel heatsink is good enough; however, I went for the Noctua NH-L9x65.

My stand is always that if you have space, opt for something bigger. It makes life a lot easier - and easier to upgrade.

I suspected exactly that after watching a Youtube video about someone building a NAS with it. The Fractal 304 is a lot easier to work with.

Definitely stop VM autorun and double check core assignments. In terms of Windows, it should be automatic but you might be asked to reactivate.

If you mean brand then as far as I know, they are all more or less same. ECC is generally classified as non-consumer-class hardware so things tend to be pretty homogeneous. As to how much RAM, I personally don't think 16GB is enough (which is why I went to the C236 chipset specifically to get 32GB). In my simplistic mind: VM (8GB) + Plex (4GB) + ownCloud (4GB) + Crashplan (8GB) = 24GB already. That leaves about 8GB for unRAID to mess about with.

I just built a server with similar parts (here). The C236 WSI does support 32 GB ECC (2 sticks 16GB each) but it has to be Unbuffered. Your link is for RDIMM (i.e. registered i.e. not-unbuffered). I used Samsung M391A2K43BB1-CPB.

You can actually pass through the GPU even if it's the only GPU on the system - you will just lose the ability to get back to the console (e.g. need to SSH in instead). It's just that nVidia GPU doesn't like it.

I would recommend getting 32GB RAM. unRAID doesn't need a lot of RAM but dockers and VMs etc all do. And trust me, it gets very very annoying when your VM is killed out of the blue because of insufficient memory. Also, check if your motherboard allow to set the integrated graphics as primary. I remember reading something about issues when passing through nVidia GPU, on 1st PCIe slot as the primary / only GPU.

Matter of time indeed. I'm planning to have strict user access so most things would be read-only and moving files are done over SSH + console. So even a nasty one will have to really try very hard to get through my swiss cheese layers. Space is a premium The 304 is going on a book shelf so it has got to be small and look good. It is also incredibly quiet, even with all the fans maxed out (my HDD during preclear reaches 41C max). My workstation is actually louder and I put a lot of pride in ensuring my workstation is quiet. An alternative could have been UNAS NSC-800 for more drive bays but (1) it only fits passive cooling motherboard and I do need a bit of power and (2) it is very very hard to get. If I had a cabinet, I would have opted for rack-mounted solutions. 6 bays give me a maximum of 40TB, using current HDD technology (8TB / drive + 1 8TB parity). That is way more than enough for me for at least medium term. I currently don't even break the 10TB mark.

After months of researching, considering options and testing, I have decided to fully jump on the unRAID bandwagon and build a dedicated NAS server. I originally considered and tested having NAS + main system in the same CPU. However, I found my existing workstation hardware is (1) not entirely suitable for an always-on system, (2) not currently fully supported by a stable release and (3) too focused on speed and less on reliability. It then became a very complicated process to pick upgrade parts which are compatible, affordable and available. Hence, I stick to Occam's razor and started to build a dedicated server. Main intended use: Ransomware-resistant NAS Plex Media Server to manage and stream my media collection Crashplan for 24/7 cloud-based off-site backup Windows VM only accessible via RDP (i.e. no GPU pass-through) Below are the parts: Case: Fractal Design Node 304 in white This case ticked all the boxes. It's compact and light-weight, yet still support 6 3.5" bays. The fans are also quiet and the case in my opinion looks good. I considered the Lian Li Q08 and Q25 but my main concern is that the fan projected airflow does not cover all bays. The 2 front 90mm fans of the 304 ensure air flows across all 6 drive bays. I'm pre-clearing 5 drives simultaneously right now and the hottest one hovers around 39-40C. Motherboard: ASRock C236 WSI There isn't many mini-ITX motherboard that has more than 6 SATA ports + integrated / on-board GPU and supports virtualisation, 32GB RAM etc. This choice is basically a trade-off to get 8 SATA ports while giving up IPMI. I considered the ASRock C2750D4I but the Avoton C2750 CPU is a little too "just good enough" in term of power. CPU: Intel Xeon E3-1245 V5 Once the motherboard is picked, this is the best value-for-money option. It supports hyper-threading (unlike the 1225) so support up to 8 threads - great for multiple simultaneous streaming / multi-tasking. It also has (free) integrated GPU so I always have access to the console. RAM: 32GB ECC (2xSamsung 16GB DDR4) A bit of an overkill but it will get utterly annoying if there isn't enough in the future. PSU: Silverstone Strider Platinum 550W It is only 140mm deep so it gives me just a little bit more space to tidy cables in the cramped Fractal 304. Review is good and it is 80+ Platinum so great for 24/7 running. I briefly considered a 80+ Titanium PSU but none is shorter than 160mm. Cache: Samsung 850 EVO 1TB This was taken from the workstation. 512GB is probably enough but why buy new when you can reuse. Parity: HGST He8 8TB I paid a bit more here for enterprise-grade hardware, given the parity drive will receive a lot of write and is critical when it is actually needed. In other words, reliability is more important than speed. Array: Seagate NAS 8TB, WD Black 6TB, Hitachi Deskstar 3TB + 4TB All but the Seagate are taken from the workstation. The last 2 are older (read: slow) but I reckon as long as they don't fail preclear, they can be reused - since everything is gonna be over the network so subjected to the theoretical 125MB/s transfer speed cap. That's it in term of hardware. Now if only preclear can run a bit faster.