May 5, 20206 yr Morning all, I get seemingly random print_req_error: I/O errors (which then causes odd docker/vm/share corruptions/issues) when the array has been on for several hours, usually at idle. Things I have tried: Swap controller cards from 9207-8i to 9305-16i Changed the SAS/SATA cable to SSD (it is on a separate cable and port from the array drives) Changed SAS controller port for SSD Cleaned SAS controller card and cable connection points Rebuilt (new docker and vm image files) and reformatted SSD Swapped SSD to previously disabled mobo SATA controller (just trying this as of right now, so system hasn't been on long enough to see errors yet) Unfortunately, onboard Asmedia SATA controller has massive speed degradation (blue line vs green/black lines in image below) compared to controller card. Booted to bare metal Win10 and attempted to run Samsung Magician (will not load, continues to say "Scanning for HDD"") Updated BIOS Been hammering away at this one for awhile. The SSD is new, I have a hard time believing it is at fault, especially since these errors reference apparently random and different sectors. The goal was to just have the SAS controller handle all of the drives and leave the mobo controllers disabled. I did not anticipate the speed drop from the controller card. Any suggestions here on where to go next? Diagnostics attached. Appreciate any and all help gents! diagnostics-20200505-0803.zip Edited May 9, 20206 yr by Skrumpy
May 5, 20206 yr Community Expert 36 minutes ago, Skrumpy said: Swapped SSD to previously disabled mobo SATA controller (just trying this as of right now, so system hasn't been on long enough to see errors yet) Unfortunately, onboard Asmedia SATA controller has massive speed degradation (blue line vs green/black lines in image below) compared to controller card. Difficult to say what exactly is the problem, sometimes the device resets, others there's a fault with the HBA, best bet would be to use the AMD onboard SATA ports, though I believe they are still from Asmedia they are not x1 like the extra Asmedia controllers, so performance should be normal.
May 5, 20206 yr Author 6 hours ago, johnnie.black said: Difficult to say what exactly is the problem, sometimes the device resets, others there's a fault with the HBA, best bet would be to use the AMD onboard SATA ports, though I believe they are still from Asmedia they are not x1 like the extra Asmedia controllers, so performance should be normal. So if this was your PC it wouldn't concern you at all that the controller was throwing those kind of errors with an SSD? To the point that it may happen with the array drives as well? FWIW, I haven't seen any errors with the onboard controller and this SSD as of today.... I thought this LSI Logic Controller Card 9305-16i was listed on the unRAID compatible list, but apparently I was mistaken.... Looks like I will have to order in an LSI SAS 9305-24i and see if I get the same issues...
May 5, 20206 yr Author So this looks like it is a TRIM issue as I have that set to run daily at 5A and it just goes haywire with the system. I was browsing the web and came across another post of yours @johnnie.black from a year ago. So the issue is that there appears to be only one specific subset of cards with a downgraded flash that supports TRIM. I wouldn't have too much of an issue (aside from spending $400 on a controller card that can't properly handle an SSD....) if the speeds on the MOBO controller were near that of the controller card. What a shame.. Quote I would say that for now, and likely for the foreseeable future, these are the options to get your SSDs trimmed: -Use the onboard SATA ports, it works with any SSD -if your SSDs support determinist trim and you really want to have them connected to the HBA get a SAS3 LSI, like the 9300-8i.
May 5, 20206 yr Author Guess I didn't do enough research in this area, really did not see that there would be a big issue with a controller handling an SSD in 2020!! So, rather than getting the 9305-24i, I've been reading that an enterprise level SSD may support DRAT/RZAT and work fine with the controller? Anyone have recommendations of something I can go with to replace the 860 PRO?
May 6, 20206 yr Community Expert LSI HBAs can only trim SSDs with DRAT/RZAT, I would expect the 860 PRO to support that, since the 860 EVO does, but never confirmed.
May 6, 20206 yr Author 6 hours ago, johnnie.black said: LSI HBAs can only trim SSDs with DRAT/RZAT, I would expect the 860 PRO to support that, since the 860 EVO does, but never confirmed. Reportedly they do. https://blog.insanegenius.com/2020/01/10/unraid-repeat-parity-errors-on-reboot/ Got new SSD errors upon TRIM this morning, going to try another cable and try the other onboard SATA controller and see if that improves speeds and resolves these.
May 6, 20206 yr Author Well, after moving the cable to the onboard AMD controller, success. At least now I know the drive isn't bad. And speeds look to be back up to where they were with the LSI controller. Not quite sure what the issue was with the onboard Asmedia controller, but maybe I have a hidden BIOS setting that needs tweaking.. or that particular port was bad.. I have an LSI SAS9340-8i on the way that reportedly does not have SSD TRIM issues like I am experiencing with the 9305-16i. I guess I'll find out on Sat.
May 6, 20206 yr Community Expert 4 minutes ago, Skrumpy said: Not quite sure what the issue was with the onboard Asmedia controller, It's a PCIe x1 controller.
May 9, 20206 yr Author On 5/6/2020 at 9:23 AM, johnnie.black said: It's a PCIe x1 controller. I thought even with PCIe x1 you could still get decent speeds depending on the protocol? Onboard controller must be running outdated protocols? Either way, I appreciate the help. I ended up going with a Art of Server 9340-8i card and I am having no TRIM issues, even after popping in another SSD and going to a cache pool. Unfortunate, but the 9305-16i is over twice as much as this 9340-8i and didn't support TRIM natively.
May 10, 20206 yr Community Expert 13 hours ago, Skrumpy said: I thought even with PCIe x1 you could still get decent speeds depending on the protocol? Max theoretical bandwidth for a PCIe 2.0 x1 link is 500MB/s, max usable bandwidth is around 400MB/s.
Archived
This topic is now archived and is closed to further replies.