Jump to content

bonustreats

Members
  • Content Count

    43
  • Joined

  • Last visited

Everything posted by bonustreats

  1. So I found out why my cache drive failed! Looks like a cheap molex to SATA connector melted and caught on fire! If any other folks see this, please learn from my mistake and use better connectors - crimped connectors or the SATA power cables are better. I posted on Reddit and a couple people gave some really great information, including a youtube review of another situation eerily similar to mine.
  2. Thanks a lot - I appreciate all the help! I'll mark this one as solved.
  3. Wow - thanks so much for the quick reply! Just out of curiosity, were you able to tell that from the server diagnostics or from the SMART report? If from the SMART report, was it because all the attribute VALUES were 100? OR was it because there were P's in the prefailure warning column? From this post it indicates that V-NAND or 3D TLC are the preferred types of SSD - is that correct? I was also looking and couldn't find a size recommendation for the cache - would 500GB be enough for...pretty much only Plex? This drive looked okay, if that's the case. Thanks so much for all the help and sorry for the billion questions!
  4. Hi Everyone, Yesterday, I noticed Plex was down and went to investigate. I tried rebooting the server, but I saw that the cache drive was unmountable. I found a couple forum posts about how to erase/rebuild the cache drive, but I'm not sure if the drive itself is bad. I ran a SMART report, but I'm not sure if the drive is dead or if it's just behaving weird because it doesn't have a proper file system associated with it. If the drive is okay and I just need to rebuild it, just wanted to confirm that this link would work for this situation. Thanks very much for your time! radagast-diagnostics-20190911-1534.zip radagast-smart-20190911-1130.zip
  5. I've had some issues before with call traces However, I got some new hardware and I'm getting them again. I looked through the log file and didn't see anything that I recognized; at least from previous posts. Are these call trace errors something to be concerned about or is it possible to ignore them? Thanks for your time! radagast-diagnostics-20171013-1713.zip
  6. Yeah, I think I'd be leery about trying a preclear again regardless - seems like you're just begging for trouble. New one should be here tomorrow. Thanks very much for your time and responses! I'll mark this thread as "solved."
  7. Gotcha. Do read failures always equal disk replacement? Is it safe to assume that any time there are errors in that column on the main screen that the drive is bad? Thanks very much!
  8. Here it is, sorry for the delay! Edit: sorry, uploaded the wrong one. radagast-smart-20170817-1237.zip
  9. I don't think I have them enabled; is this the correct way to enable them?: https://wiki.lime-technology.com/Configuration_Tutorial#Install_UnMENU_for_Screen_and_Email_Notifications Extended SMART test is currently running, but I think it said to allow 2 hours per TB. I can just download the results and paste them here, correct? Thanks for responding!
  10. Gotcha, I was guessing/hoping it's something along those lines. Write activity is suspended. Thanks!
  11. Hi Everyone, Setup: 1x4TB WD Red (parity), 2x2TB WD Green, 2x2TB WD Red, 64 GB SSD Cache, 2GB Flash. On the Main tab, under the "Array Devices," unRAID is saying that there are 239,435 errors for the parity disk. However, under the Parity History, the last check completed with 0 errors (run on 14AUG) - please see attached image. Is this something to worry about? I tried searching the forums, but could only really find info about actual parity errors that come up during a check. Diagnostics are also attached. Thanks in advance! radagast-diagnostics-20170816-1803.zip
  12. Meowlrighty, here they are - thanks for the quick reply! radagast-diagnostics-20170301-1835.zip
  13. I got an update to FCP, and ran a rescan - it's now showing the irq 16: nobody cared in the GUI, and it's telling me to post my diagnostics. Will posting the diagnostics now provide any new information compared to the one from Saturday? If so, I'll gladly post them. I haven't seen any particular issues of note; maybe some transfer speed reduction in some cases, but that might be related to the hard drive that's being used for that copy.
  14. I guess it hasn't, because it is now back again; same "irqpoll nobody cared" error. The question now is: is this a serious problem or just some kind of background process that's not going to hurt anything? radagast-diagnostics-20170225-1151.zip
  15. It looks like other people are running into the same issue that I was (call trace errors, at least), and unfortunately, I am again. I got a new (to me) MB, processor, and RAM from a friend of mine early last week. Swapped everything over, and it booted right up, I couldn't quite believe how easy everything was. I went through and updated all my plug-ins, as well as went from 6.2.4 to 6.3.1. I ran into a bit of a snag trying to upgrade to 6.3.2, but figured it wasn't a big deal and would upgrade later. Parity check ran; everything seemed great again. Unfortunately, the same call trace error popped up in Fix Common Problems again. I ran the diagnostics to post this yesterday, but it seemed like the forum was down or something, so they're here now. I didn't do a windiff or anything, but it looks like the exact same error that came up before. I was able to upgrade to 6.3.2 and (so far) the issue hasn't resurfaced, so I'm not sure how useful the diagnostics are. However, that makes me wonder how the heck the same error could crop up with completely different hardware. Could it possibly be my flash drive? radagast-diagnostics-20170220-1301.zip
  16. Ha - I wouldn't ever again either, now. That thread said they got out of the mobo game in 2013, so hopefully you CAN'T in the future I've heard mixed things about Gigabyte, but (so far) no complaints about Asus, so I'll probably look into a replacement Asus board. Thanks for the support - it's pretty aggravating when stupid things like this crop up. Should I mark this thread as "Solved"? Or maybe "Closed" or something? Thanks everyone for your help and input!
  17. Just to double check: the unRAID OS is agnostic to the hardware (for the most part), right? If I switch the hardware (mobo, processor, RAM, etc), unRAID won't even notice, as long as it's booting into the same flash drive/hard drive configuration as before, correct?
  18. Well, I tried the cable...no dice. Did some more googling, however, and found this thread: https://communities.intel.com/thread/30813 It basically describes my problem and troubleshooting attempts to a T (my board is DH67CL). It turns out that you can brick your motherboard if you 'jump' too many BIOS update versions. Not sure if this is an Intel-only problem, but I tried to go from version 105 to the newest one, which is 160. I saw NO disclaimers on the Intel website warning about this problem, so if anyone uses Intel boards for their unRAID builds, this would probably be useful information to give to them. Not sure where that should go, though...should I email an admin? I may be able to "unbrick" it by using a next generation (compared to mine) processor, but I don't think I want to buy a processor just for a potential fix, only to find that it doesn't work and now I'm out the money for the processor, too. I guess the hunt is on for a new motherboard, at least. Unless now is a good time to upgrade other internals...
  19. Thanks - I borrowed a DVI to VGA cable today, so hopefully that theory can be tested. I do have a video card and can try that next. If that doesn't work, I'll try to google a little bit more, in case there's something I missed. Otherwise, it might be new mobo time. I'll try the cable and all that jazz later this afternoon and will report back.
  20. Sorry for the delay in reply - work was a little crazy yesterday, so I just got to this today...aaaaand I've run into an issue. I ran the F7 BIOS update from here: https://downloadmirror.intel.com/22273/eng/BIOS%20Update%20Readme.pdf. I double checked the correct BIOS version and put the .BIO file onto a recently formatted 1GB flash drive. I went through the utility, I got a message saying something like, "BIOS update successful, rebooting machine." Only...nothing came back up, just a black screen with a "cable not plugged in" message. I tried restarting multiple times, did a fair amount of googling, and settled on following Intel's advice (http://www.intel.com/content/www/us/en/support/boards-and-kits/desktop-boards/000005753.html). I tried Option 1, but the computer didn't automatically boot into Maintenance mode; so I tried option 2. Pulled the CMOS battery for ~20 minutes and tried again. Same result - no screens loaded. Then I tried the BIOS recovery option (http://www.intel.com/content/www/us/en/support/boards-and-kits/000005630.html). Still nothing. The only thing I can think of is that the onboard HDMI is not the default output, and that the onboard DVI is, resulting in no screen image and all kinds of insanity being inflicted upon the motherboard through my multiple 'failed' attempts. Unfortunately, I don't have a DVI-to-HDMI cable to test this theory, even though it sounds weak to me. Why wouldn't the board output to either port? I'll try to borrow a cable tomorrow to test. Has anyone run into a problem like this? I'm really lost and I'm REALLY worried that I just screwed my board and my server. Thanks, Jeff
  21. You could try the unBalance plug in. I used it to convert my disks to ZFS. https://lime-technology.com/forum/index.php?topic=45352.0
  22. That matches the owner of IRQ 16: So the problem is likely to affect the boot device. So does that mean that all USB ports are affected or just that particular one? If I switched the unRAID stick to another USB port would the problem go away or is this indicative of problems to come (like if one USB port goes bad, they'll all eventually go bad)? I couldn't find any of my flash drives to upgrade the BIOS last night, so I'll try to grab one today and update it later this evening. I'll then restart the server and see if the problem comes back again. Should I add any other steps?
  23. Thanks for the info! I'll try the BIOS update tonight or tomorrow. I don't really have experience with unRAID RCs and a little leery about RC software in general, so maybe that'll be a last ditch effort. If the BIOS doesn't fix it, I'll try the irqpoll option. Speaking of fixing, is this a persistent error (once it happens it 'stays on' or can it occur at any time? Would a server restart clear it from unRAID?
  24. Computers are very effective air cleaners. You don't even want to touch the inside of a tobacco smokers computer. I try to clean my hardware 2x a year, once in spring, once in fall because I have two things keeping my apartment dust filled: forced air and a cat. I do that ever since a friend of mine gave me some of his old hardware (which I used to make my first unRAID build) and it was pretty freaking gross (please see attached pic). I can't even imagine a smoker's machine - yuck.
  25. IRQ 16 is used by a USB 2 controller: This is where I struggle, I confess, trying to link the controller via a hub to a specific device. So I can't say what the device is. An obvious candidate would be your USB boot device, but I just can't say with any certainty. Maybe someone else can help here. One thing I do notice is that your BIOS is dated 2011 so it might be worth looking to see if an update is available. Failing that, maybe try the "irqpoll" boot option that's suggested. I'm not sure why you think your CPU might be overheating. I can find no evidence of that. Thanks for looking! Do you mean that the boot device itself is potentially bad or the port to which it's attached? I just looked and there is a BIOS update from 2012, so I can try to apply that when I get home later today. I tried to look up what irqpoll is, and a lot of it goes over my head (sorry); however, it seems like when this is thrown, it indicates potential hardware failure. Not sure which component, of course, haha. In another thread (https://lime-technology.com/forum/index.php?topic=35590.0), they said that the problem was CPU overheating. I didn't see any evidence when I looked through the logs, but figured it might be something good to do anyway and to reduce the number of potential variables. I'm pretty woefully ignorant on log perusal, so if you don't mind me asking, what do you look for when going through a log?