May 30, 20188 yr Community Expert Okay so run the command and terminal waiting for a long period of time is normal. Yes, IIRC it had about 200GB to move.
May 30, 20188 yr Author Is it okay if I restart the server then re-run that command you gave me... Looks like another stick of ram is acting up? I noticed when the sel starts reporting the ram issues cpu usage spikes to 90% sel4.txt
May 30, 20188 yr Community Expert Wait for the command to finish, as long as the errors are being correct it should't make much difference.
May 30, 20188 yr Author Heyyyyy it finished that worked. You are the current MVP! I gave the company a call they recommended trying to reseating the cpu so I am going to give that a whirl on the next error. Edited May 30, 20188 yr by taserz
May 30, 20188 yr 15 hours ago, taserz said: Okay so I just realized I can expory SEL from IPMI here is what it says sel.txt Curious that the log shows two errors every second from start to end. Edited May 30, 20188 yr by pwm
May 31, 20188 yr Author 10 hours ago, pwm said: Curious that the log shows two errors every second from start to end. Yeah that seems correct for each of the sel logs I have seen. Do you think it is a bad board? I have reseated cpu's and that didn't change anything
May 31, 20188 yr 5 hours ago, taserz said: Yeah that seems correct for each of the sel logs I have seen. Do you think it is a bad board? I have reseated cpu's and that didn't change anything It's a very, very strange behavior because RAM errors are extremely stochastic and so normally have very random timing patterns. If the RAM has an absolute bit error, i.e. 100% repeatable, in a memory cell that is used by an action that is performed twice/second, then it's natural to get the error every two seconds. But if you reboot and still get the same interval, that is very strange. All modern OS uses a memory mapping layer and would normally not manage to map the same physical address range to the same software routine or data. Obviously it could happen, if there is some timer-based logic that gets mapped very early in the boot sequence. One example is the code that handles uptime or real-time-clock in the kernel - it starts counting time extremely early. And that would explain why the repeat frequency is so well synchronized with the time. But still extremely strange that a machine with gigabytes of RAM would manage to have a persistent bit error within the very few bytes of memory that happens to be needed by some tiny core timing logic in the kernel. And another thing - if booting some standalone memtest software instead of the normal unRAID kernel it should be virtually impossible to get the same memory access pattern logged in the event log. I definitely think you should try to switch around the modules and figure out what happens with reported fault address and fault frequency. Also try to download a separate bootable memtest image. Just so this isn't instead a false memory error caused by a software bug somewhere.
May 31, 20188 yr Author So I pulled another stick of ram from cpu1 that never reported any issues and seated it in the same slot that was throwing errors for that other stick of ram on cpu2. If it throws an error with a known good stick and that same slot would we assume it is a bad board?
May 31, 20188 yr 1 hour ago, taserz said: So I pulled another stick of ram from cpu1 that never reported any issues and seated it in the same slot that was throwing errors for that other stick of ram on cpu2. If it throws an error with a known good stick and that same slot would we assume it is a bad board? If a "known good" module gives error in that slot, then something is wrong with the board or the firmware responsible for handling ECC events. It's easy to forget that there is a separate computer system responsible for the hardware supervision and remote maintenance.
May 31, 20188 yr Author Yeah I kind of figured that as well. I am assuming it's a bad board but I am not excited about waiting for a new one. To most dual socket mobo's allow me to run it off one cpu? Figure I could do that in the meantime.
May 31, 20188 yr 13 minutes ago, taserz said: To most dual socket mobo's allow me to run it off one cpu? Figure I could do that in the meantime. Yes, with caveats. Half your memory slots and maybe some of the PCIe slots may not work.
June 1, 20188 yr Author I have all the stick of ram on CPU1 and pulled the second CPU. TheServerStore is sending me another board so hopefully, this solves the issue. I will keep everyone posted.
June 6, 20188 yr Author Just got the replacement motherboard into today going to try to give it a whirl again.
June 30, 20188 yr Author I want to let everyone know I got the new motherboard and it's a breath of fresh air! It is running smoothly. The last question I wanted to ask was how would I use my 2x 500gb ssd as 1tb ssd cache pool aka stripped.
June 30, 20188 yr Community Expert 4 hours ago, taserz said: The last question I wanted to ask was how would I use my 2x 500gb ssd as 1tb ssd cache pool aka stripped. You can, see here for the possible modes: https://lime-technology.com/forums/topic/46802-faq-for-unraid-v6/?do=findComment&comment=480421
Archived
This topic is now archived and is closed to further replies.