allanbennett

Stopped making reflinks and haven't had an issue since. arcstat shows my max cache memory set to 12G. Not sure how unraid determines what this value should be but for 40TB this seems low. With arcstat and zpool iostat, I now have the tools to do further testing but it wont be this week. If I find any setting that allow me to use reflinks again, I'll update my post. Thank you Michael_P for your help!

This is taken now, I can supply again next time it happens. What are you looking for? ps_auxf.txt

I do transcode to my cache pool and my previous oom was when there was nothing streaming on plex. But sure I can set limits on all my dockers. Did I mention I have another unraid server with similar dockers? It has regular array with xfs drives (rather than zfs pool) that has never had oom issues.

I agree that something starts using huge amounts of memory. I suspect its zfs as no process shows abnormal memory usage when this happens. The unraid dashboard page doesn't update when this happens. Is there a way to see/track zfs cache? I will look to see if I can set a zfs cache limit but i have read it can be ignored. When I created zfs pool I choose raid z1 and compression. Very little on the disks is compressible but i read it would still be a performance boost. Any zfs experts out there? My next step is to stop creating reflinks and possibly removing the ones I already have. Plex transcode segfaults while not common are not uncommon either. My servers see one or two a week. That segfault happened about 20 min after the server returned to normal. At the time of the lockup I had 12 plex streams playing.

I think it just did it again but this time no oom but server locked up for a good 15 min. proxy-manager still off. root@unServerD:/boot/config# free -h total used free shared buff/cache available Mem: 94Gi 86Gi 2.2Gi 198Mi 6.7Gi 7.5Gi Swap: 63Gi 9.2Gi 54Gi I was able to get swap working. Can't use swapfile on zfs so the plugin swapfile plugin doesn't work. Here is what I did in case it helps others: https://askubuntu.com/questions/1468817/how-to-solve-error-on-swap-creation-skipping-it-appears-to-have-holes-ubu tldr; /dev/zvol/cache/swapfile none swap discard 0 0root@unServerD:~# zfs list NAME USED AVAIL REFER MOUNTPOINT cache 446G 454G 378G /mnt/cache zfs create -V 64G -b $(getconf PAGESIZE) -o logbias=throughput -o sync=always -o primarycache=metadata -o com.sun:auto-snapshot=false cache/swapfile mkswap -f /dev/zvol/cache/swapfile added this is go file swapon /dev/zvol/cache/swapfile unserverd-diagnostics-20251009-2119.zip

Definitely worth a try. proxy-manager off. I am also removing unused dockers and seeing if i can get swap working. I shouldn't need it, but if the issue is zfs, I've read this can be a work-around.

Funny you suggest this. This docker has always had --memory=1G --no-healthcheck. I use this for cert renewal. This docker can be off for a month.

Just oom'd again about an hour after reboot. tdarr docker was removed before reboot. Plex might have had 1 or 2 people connected but probably not. I had htop running load went over 200 before it became unresponsive. tail on syslog shows me oom messages after waiting about 20 min. ssh sessions basically unresponsive for last 30 min. Remote console loging times out. Ill give the system a hour to see if I can get the latest diagnosis. well only took another 5min as I as writing this up. Only thing different with this server compared to other server is zfs pool. I wanted to use reflinks. is this a possible culprit? unserverd-diagnostics-20251009-0843.zip

I can certainly try this. I tdarr/ffmpeg had been complaining about out of device space and encodes would fail. So I had not been sending jobs to it for "about a week." I was only able to fix it by uninstalling tdarr and deleting appdata files. So tonight was first time in tdarr had been running transcodes when server oom'd. I also had this problem on another of my unraid servers very similar specs as this but I haven't had any oom on it. Plex and tdarr still running this morning, qbittorrent still non-functional after it was oom killed. Rebooting server now with tdarr removed. Let's see if it can go through this evening without oom'ing. I capture memory usage right after oom and then about 5 min later. I wasn't able to tell what was using the memory between the two, root@unServerD:~# free -h total used free shared buff/cache available Mem: 94Gi 91Gi 1.8Gi 787Mi 2.7Gi 2.7Gi Swap: 0B 0B 0B root@unServerD:~# htop root@unServerD:~# free -h total used free shared buff/cache available Mem: 94Gi 16Gi 44Gi 782Mi 35Gi 78Gi Swap: 0B 0B 0B Do i need to setup swap manually to get it to work?

I'm sorry if this should be posted in else where as I am running 7.2. I had been running 7.2.0-beta.2 for about a month with no problems until about 5 days ago. I have tried upgrading to RC1 same issue. Memory usage seems to increase very quickly. I wasn't able to determine what was using the memory. I tried install the swap plugin but my cache drive is zfs not btrfs so maybe it wont work, it doesn't show any usage. Dell R720 Processor Intel® Xeon® CPU E5-2690 v2 @ 3.00GHz Memory: 96 GiB DDR3 Multi-bit ECC Intel GPU Arc A380 zfs storage, cache pool, no array unserverd-diagnostics-20251008-2137.zip