BobPhoenix

Since the data is not STRIPED on the disks just use a bootable Linux CD and you should be able to access your data. Also you should email lime tech directly as you would likely get a faster response that way. But like I said if you use a bootable Linux CD/USB Flash you should be able get to your data while you wait for a response to the email.

Correct. My ASRock - EP2C602-4L/D16 has a Marvel 4 SATA port controller that I do not use at all. The other 10 SATA ports are from the Intel chips and work fine both for the array and passed through to VMs. I would do like you said and transfer all drives to LSI controllers and have nothing connected to Marvel controllers. My guess (and it is a guess since some people do not have problems with Marvel controllers) is that will fix your access problems. It is suggested it is a combination of drive manufacturer and SATA interface version on the drive that causes it. I haven't found a combination that works for me except switching controllers to LSI. If I had more time to experiment I might be able to come up with something else but swapping controllers is just so much easier.

Beware Marvel controllers and unRAID v6 however. You might be one of the lucky ones where you won't have any problems with Marvel controllers but you might not. I say this because my ASRock - EP2C602-4L/D16 has 14 SATA ports but 4 of them are on a Marvel MB controller and I had dropped drives all the time. Luckily I was passing it through to a VM so I never had problems with dropped drives from my array. But it was annoying having to reboot the server every time it happened in order to get access to the drives in the VM.

When I RMA'd a drive to WD one of the choices was "drive access is sluggish" in other words nothing obviously wrong showing in a smart report. So I don't think you have anything to worry about.

I would get 2 8TB drives and store one at a relatives house. Then as you add photos you swap your drives with the relative so that you have backups at an off site location. You might even consider a cloud backup as well for a third off site location. If I had any pictures that precious to me I wouldn't take any chances.

Didn't think about that very well. SATA isn't SAS but SAS can use SATA devices.

My take on why SSDs are not recommended in the array is that Trim would not work on an array drive because if you did use Trim it could invalidate the parity protection.

Figures.

What linux kernel is the QNAP using? I know 32 bit unRAID has no problem that I ever saw with Marvel controllers. But 64 bit unRAID would drop drive on me all the time.

They also make PCI/PCIe cards that contain a com port - you could try that as well.

I knew he was in prison but I thought he was convicted in California and was serving time there?

Yep I have the following cold spares precleared and ready to insert in my unRAID arrays (note the plural there): 8TB WD Red non-pro, 4TB HGST cool spin (non-NAS), 6TB WD Red non-pro, 5TB Toshiba, 6TB HGST NAS & 4TB WD Red non-pro. This way I don't need to shut down for an extended period just minutes. I just pop in the replacement and buy a new drive to preclear and become the new cold spare. The 6TB WD Red non-pro is a waist at this point because that server is in the process of being converted to 8TB WD Red drives. But I have the 6TB available from when it was composed of 6TB and 4TB Red drives. The 4TB drives were replaced with half that many 8TB drives when I bought 8 drives at once last May. The only array drives I don't have a cold spare match for on my servers is 3TB toshiba. Those could become 5TB toshibas potentially but I'm really thinking I might just go to 8TB HGST NAS drives instead.

If you have access to v5 you could upgrade 4.5.6 to v5 and then you could upgrade to 4TB drives. If your computer supports 64bit then I would recommend unRAID v6 as v5 is not downloadable any more so you would have to contact lime to get access to it. The download page says to get version prior to 6.2.4 to do the following: " If you require a previous release, please send a request to [email protected]."

Personally if I were to get into this situation I would shut down my server until I had a new drive tested and ready to add as a replacement. One reason I try to never be without an empty replacement drive on hand to use.

Wouldn't the MB connector (SFF-8087 in my GEN7 anyway) work with a SAS expander? I've never tried it myself with my RES2SV240 but would be an interesting experiment. Don't really expect it to work but thought I would ask those more knowledgeable about this.

I would turn of the ACS overide if you want all of the nics going to the same VM. Also you need to stub the device in the syslinux configuration file by editing it from the flash configuration (click on "flash" on main tab in unRAID GUI) and add it's device ID. So to get a nic on my VM server I have this defined like the following: label unRAID OS menu default kernel /bzimage append vfio-pci.ids=8086:1d6b,1131:7160,8086:10d3,1912:0014 initrd=/bzroot label unRAID OS GUI Mode so for you enter the follow device ID where I have my list: 8086:10a4 So you might have: append vfio-pci.ids=8086:10a4 initrd=/bzroot You will probably have other information on the append line so don't remove that just add the vfio-pci.ids=8086:10a4 to it or if you all ready have the vfio-pci.ids= present then just add the nic device id (8086:10a4) after a comma like in my example above.

Because you have 4 nics in 2 IOMMU groups you will have to pass them through in pairs to a VM. Unless you can separate each nic into its own IOMMU group by using the "PCIe ACS Override" in the VM Manager settings. The other problem you might have is that those IOMMU groups have "bridges" in the same IOMMU groups and KVM isn't able to pass through devices behind a bridge - at least for I've tried anyway.

The LSI 9201-16i can handle 16 drives at a time. But LSI cards (of the 9201 vintage anyway) do not allow the TRIM operation (you do TRIM from unRAID not BIOS) to work correctly unless you have Enterprize SSDs with a specific capability. @johnnie.black could give you the exact name and it is posted on the forums I just haven't had time to find one of his posts that have it. Asmedia cards don't have the Marvel bug so they might be an option but they may not support trim either. Maybe @johnnie.black could tell you if they (Asmedia controllers) support TRIM. If you do decide to get a -16i and want cheaper prices then try eBay but go for NON-China seller and a used server pull. There have been reports of the China ones being counterfeit boards - why I have purchased server pulls from US sellers on eBay. If I still haven't answered ALL of your questions let me know what I still need to answer and I will attempt to answer them. But maybe tomorrow before I get to them.

Use MB connectors for SSDs so that TRIM works on them it doesn't unless you have a option that is usually only available on Enterprize SSDs with LSI controller. Array drives use MB or LSI controller.

Check this thread: https://forums.lime-technology.com/topic/59091-marvel-issues-starting-point-for-investigation/ and this one: https://forums.lime-technology.com/topic/39003-marvell-disk-controller-chipsets-and-virtualization/ For me the drives would drop at random intervals. If I'd had them as array drives I would have been forever running parity checks to recover. Luckily I had my Marvel controller going to a VM so it was only a nuisance to that VM and the fact to recover from it I had to restart the whole server not just the VM. There are work arounds that are supposed to help one of which is to turn off virtualization in the BIOS of the MB which obviously wouldn't work for me. The recommended controllers are LSI based like the Dell H310 controller which has to be cross flashed to an LSI 9211-8i. I use LSI 9201-16i's as they don't need to be flashed. The -8i version also does not need to be flashed and therefor is plug and play with unRAID.

That controller mentions "Marvel" in the description. Marvel controllers are NOT recommended with unRAID v6+ servers you could have problems but may not. They work fine with 32 bit unRAID v5 servers.

If it was Windows VM I would use RDP and VNC for linux. Not sure why TightVNC doesn't work for you - always works for me - sorry.

What are you using for a port number in TightVNC? I use ports starting with 5900 (the default) so for instance to connect to a VM on my VMServer I use this in TightVNC to connect: VMServer::5900 I look up the correct port with noVNC built into unRAID and adding 200. The port number for the same VM above with the built in noVNC is 5700.

Do you access the same relatively small (I.E. not large video media files) on a regular basis (I.E. multiple times a day). If so then the SSD portion may help you. Otherwise I would suggest you spend more and just get an SSD of sufficient size or use a standard spinner and save money. Those hybrid drives work better on an OS drive where the parts of the OS that get accessed the most get the boost from the small SSD cache installed on the drive. Or on a gaming rig where portions of the game that get accessed the most get loaded in the SSD cache. Based on the specs at Seagate that SSD cache is only 8GB so would only help with smaller files like I thought. https://www.seagate.com/www-content/product-content/firecuda-family/firecuda/files/firecuda-ds-1903-1-1606us.pdf

I always run 3 cycles of preclear even on my 8TB drives. I've had drives die on the third cycle before - none of the 8TB drives - but I have had others (WD 2TB EARS) work perfectly on the first two cycles and die on the third. Preclear time: 120:17:18 for 3 cycles on a NON pro WD red 8TB - so yes takes a LONG time.