considerthecricket

I managed to kind-of catch the issue this morning before it rebooted or had a panic attack. I specifically logged into the CLI from console last night and left it up. I do agree it is an OOM issue but the logging was skewed in some fashion, backlblaze backup according to today's output spikes to consuming a whole crap ton of ram sometime during the night (7.5Gb worth) and isn't playing nicely. Example from logs (truncated a bit) Dec 16 06:20:47 Tower kernel: [ pid ] uid tgid total_vm rss rss_anon rss_file rss_shmem pgtables_bytes swapents oom_score_adj name Dec 16 06:20:47 Tower kernel: [ 570766] 0 570766 9772504 4346299 4344553 0 1746 35758080 0 0 qemu-system-x86 Dec 16 06:20:47 Tower kernel: [1569864] 99 1569864 2872730 1953259 1953246 13 0 22814720 0 0 bztransmit64.ex Between the VM and backblaze, its an ass kicker.

I had suspected that too, but I think ffprobe is a symtom of the root cause, especially considering my base outage began hours before the first OOM error percolated to the logs. I've got an 18hr uptime with the same approximate set of things running now with parity check still chugging along. I know with virtualization, the processes are abstracted from the hypervisor, but is there any chance some of these logs are scraped from the guest as well? If so, ffmpeg could be causing some issues with this dump considering when it lost its mind I was streaming a game from sunshine running within an ubuntu VM. But it's got a locked set of RAM at the guest level and is told "No more" beyond that.

I have had C state 6 disabled since I stood the box up in 2019, I had no joke a 200d uptime running 7.1.2 before this began with near no warning.

Over the last 3-4 days, my unraid box has arbitrarily soft locked itself. WebUI ceases to function SSH will not reply Responds to PING None of the containers work externally The VM On it pings but is non-responsive. Terminal operations are incredibly slow to print but otherwise CANNOT log in as login times out after 60s. I managed to get it to recognize things were failing (I guess) and it dumped the diag logs and rebooted itself after ... 3-4 hours of dumping the logs. I've gone ahead and said "Eh, maybe its the version" and updated to 7.2.2 as troubleshooting, additionally I downgraded the GPU drivers from 590 to 580 (nVidia card for GPU passthru) -- There has been no tangible difference in results with these changes. Last night it froze while I was using the VM, no other container was under heavy load (See: Plex or Jellyfin had no users that I'm aware of, audiobookshelf is not resource intensive). I had nearly a 200d uptime with this sucker before this began and now I struggle to hit 20hrs. In the logs there are some OOM errors but this appears to be DURING the dump window, not before it locked, it locked approximately around 10P EST -- Specifically the entry at 21:59:07 was me attempting to connect to it because I realized it was running like a pile of dung. Any assistance would be WILDLY appreciated, my next step is an aggressive MEMTEST session or restrict WAM to the VM and see what happens. p.s. thanks to the discord crew in Support for the assists, helped me go in the right direction to find it. tower-diagnostics-20251215-0105.zip