November 9Nov 9 Author IS it not exactly my issue ?https://forums.unraid.net/topic/199499-moving-internal-emmc-boot-drive-to-a-new-nvme-drive/?utm_source=chatgpt.comSame emmc drive, coming also from a steam deck
November 9Nov 9 Yes, I think that is very likely your problem. It is the same module, and I have seen the same pattern before. Another user here had a FORESEE 64GB module like yours, from a Steam Deck. They had random crashes in unrelated programs, memory damaged to zeros, and smartctl crashing again and again. Their memory tests passed. After they removed the module, the server ran a full week with no crash, where before it crashed within a day or two. On your server, smartctl was also one of the programs that crashed most often, and the problems started about 3 days after you moved to the module.14 hours with no crash is a good start, because both RAM pairs failed within about 8 hours. But it is still short, and the test changed several things at once. Please do this:1. Confirm that the module is physically out of the server, not only unused. If it is still installed, shut down and remove it.2. Change nothing else for 3 or 4 days. Stay on 7.3.3, keep the four sticks and leave CoreFreq as it is. They were all there before Oct 25 and also during the crashes, so a clean run still points at the module.3. Post diagnostics after that, or before you reboot if anything crashes. With the log, I can confirm that nothing crashed without anyone noticing.If it stays clean, do not use that module again for boot or for any storage in this server. You can stay on the USB stick, or move to a different boot device later. After that you can update Unraid as usual.Then the cleanup:- Reformat the cache as planned and restore from your Appdata Backup.- Data written between Oct 25 and Nov 8 may be damaged, even where scrub is clean, as happened with Radarr. If an app shows database errors after the restore, restore that app from its own backup from before Oct 25.- You can also go back to a Docker vDisk instead of the directory. The earlier docker.img errors were most likely caused by this same problem.If something crashes again with the module removed, it is not the cause. Then the CPU, board and power supply are next.
November 11Nov 11 Author Here we are after (almost) 72 hours (2d and 21 hours to be exact :P)Absolutely no issue, glances and seer still up and running, no weird log in my containers, nothing obvious at leastHere is the diagIf it's ok I want to upgrade now (to 7.4.0), stay on the usb for now, and about the cache, I have to buy a 1to nvme for the cache because i'm anticipating my computer will still evolved, so I will do that in the next few days on a brand new nvme device.I will continue to track any problem regarding corrupt data, but so far so goodtower-diagnostics-20261111-1812.zip
November 11Nov 11 Thanks, these diagnostics appear to confirm it. In almost 3 days there is no crash of any kind. There are no segfaults, no kernel errors, no Docker errors and no Btrfs errors. Before, the server crashed about every 8 hours, with all four sticks, with each pair, and with or without CoreFreq. Now all four sticks and CoreFreq are back, and the only change that matters is that the module is gone. The diagnostics also show that it is not detected at all.So the cause was that 64GB module, the same as in the other thread. Your RAM is fine, and you do not need the RMA. Please keep the module out of this server for good, and do not use it for boot or for storage.Updating to 7.4.0 now is fine, and so is staying on the USB stick. If anything crashes after the update, post diagnostics before you reboot.For the new NVMe cache, I suggest this:1. Run Appdata Backup again, with the destination on the array.2. Disable Docker and the VM service in Settings > Docker and Settings > VM Manager. Mover does not move files that are in use.3. Set the shares that are on the cache (appdata, system, domains and any others) to primary storage cache, secondary storage array, with mover action cache to array. Then run mover and check that the old cache is empty.4. Create a new pool with the NVMe. Change those shares to primary storage on the new pool, secondary storage array, with mover action array to the new pool. Run mover again.5. In Settings > Docker, change from a directory to a Btrfs vDisk on the new pool, then reinstall your containers from Apps > Previous Apps. Your settings are kept in appdata.6. Enable Docker and VMs again and check that everything works.When you no longer need the old SSD, format it before you use it again. It is healthy, and the format also removes the 8 scrub errors, which only exist in the current filesystem.Keep watching your apps as you said. Data written between Oct 25 and Nov 8 can be damaged even when scrub is clean, as happened with Radarr. If another app shows database errors, restore it from its own backup from before Oct 25.If any crash comes back after you install the new NVMe, post diagnostics before you reboot.
November 11Nov 11 Author vdisk is better than directory ? because I don't mind staying as directory unless it's better to go back to vdisk
November 11Nov 11 I recommend vDisk because I've seen some weird issues with Docker directories, but many users use that without issues, so if you prefer it, it's fine.
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.