Jaytie

Hi @ich777 , tpill90 released an update for steam-lancache-prefill which contains a fix for an issue with downloading Battlefield 6. It would be really great if you could merge the newest version 3.4.0 to your container! Thank you!

To be protected agains Ransomware I execute this awesome script from a second unraid server, which has read-only access to the files on the main server. So the backupserver "pulls" the files. Due to some changes in my config I'd like to only use my main server with one or more external hdds. But if the server has 24/7 access to the backup destination where the files are pushed to, there's a higher risk of ransomware attacks affecting also the backup. So can this script be modified to manually mount the external usb disk, do the backup and unmount the disk when finished? That might solve the problem. Best wishes!

Hello everyone! I'm using the lancache-prefill container, really useful! Is there a way to implement an email-notification with the content of the logfile, maybe with some keywords? I'd love get an email notifying me, when there were new updates the cache was filled with. Greetings and thanks in advance!

All I did was a short test in order to find out, if my problem was sunshine related. No tweaking done or anything. That said, Parsec was a little bit stuttery. Quite happy with Sunshine and Moonlight right now. Since it's open source, I also prefer it. But I might have to do another test, since I'm running the newest drivers now.

I'm still struggling here and I'm also not quite sure if this problem is related to SR-IOV. But I was able to pinpoint it litte further. It doesn't matter what software I use (Parsec, Sunshine/Moonlight, RDP). So for example Sunshine/Moonlight: I create a virtual display with VDD which is used by sunshine. I connect with moonlight and so far everything works and sunshine uses the virtual iGPU (00:02.1) to encode the stream. (I can see it is utilized) The problems starts when I try to launch a game, which uses the iGPU too. Game is running but without picture. When I go to Windows/Settings/Display/Graphics and assign "GPU2: Microsoft Basic Render Driver" to that game, it works, but with unusable performance obviously. So it seems like, when sunshine + game use the iGPU there no picture (the game still utilizes the iGPU, though). Is this expected? I'm stuck. Can someone help here? edit: Well, I think I can confirm now, that this has nothing to do with SR-IOV. After almost two days of digging, I started from scratch and checked some things that I thought are already fine: I deinstalled and Intel GPU driver and installed the most recent one and couldn't believe it, but now it seems to work. So ist was a driver problem all along. Well, lesson learned i guess.

Yes I use the same one. Got it working, kind of. Normal Desktop Mode works while encoding with h.265. But when starting a game theres only sound and a blank white/black screen. All other virtual displays are disabled. It's weird because it works while on desktop but every game I have tested has no picture (windowed or fullscreen) but uses the vgpu as expected...

right 🙃 So are you using an IddSampleDriver instead of dummy plug? will do that! Thank you!

Hello everyone! First of all thanks a lot for providing this awesome plugin! I successfully use this plugin to provide a vgpu to my win11 VM. Starting a game utilizes the GPU as expected. I also wanna use sunshine/moonlight. I hope sunshine can use the vgpu to decode with h264/h265 or av1. Since sunshine requires a connected display and I cannot "connect" a physical display to a vgpu using sr-iov, I was wondering if installing a virtual display does the trick. What's required to use sunshine with the vgpu using sr-iov? Advice is much appreciated! CPU: i5 12400 GPU: UHD 730

This workaround fixed it for me. Seems to be a known permission issue:

@ph.neutral: any luck with this? I'm facing the same error.

One addition: After restarting my jellyfin container I stumbled across this warning message, that I didn't recognize until now: WARNING: Your kernel does not support swap limit capabilities or the cgroup is not mounted. Memory limited without swap. Is this normal behaviour if I use --memory=16G ?

Alright, thank's a lot for your help!

Strange, that seems to be the case. Thats the log entry when the docker memory limit is reached: Jul 16 11:01:14 Homebase kernel: Memory cgroup out of memory: Killed process 11551 (ffmpeg) total-vm:3747300kB, anon-rss:1277744kB, file-rss:0kB, shmem-rss:0kB, UID:0 pgtables:4276kB oom_score_adj:0 Jul 16 11:01:56 Homebase kernel: av:hevc:df0 invoked oom-killer: gfp_mask=0xcc0(GFP_KERNEL), order=0, oom_score_adj=0 Jul 16 11:01:56 Homebase kernel: CPU: 7 PID: 16827 Comm: av:hevc:df0 Tainted: P O 6.1.74-Unraid #1 Jul 16 11:01:56 Homebase kernel: Hardware name: Gigabyte Technology Co., Ltd. B760M DS3H DDR4/B760M DS3H DDR4, BIOS F16 12/14/2023 Jul 16 11:01:56 Homebase kernel: Call Trace: Jul 16 11:01:56 Homebase kernel: <TASK> Jul 16 11:01:56 Homebase kernel: dump_stack_lvl+0x44/0x5c Jul 16 11:01:56 Homebase kernel: dump_header+0x4a/0x211 Jul 16 11:01:56 Homebase kernel: oom_kill_process+0x80/0x111 Jul 16 11:01:56 Homebase kernel: out_of_memory+0x3b3/0x3e5 Jul 16 11:01:56 Homebase kernel: mem_cgroup_out_of_memory+0x7c/0xb2 Jul 16 11:01:56 Homebase kernel: try_charge_memcg+0x44a/0x5ad Jul 16 11:01:56 Homebase kernel: charge_memcg+0x31/0x79 Jul 16 11:01:56 Homebase kernel: __mem_cgroup_charge+0x29/0x41 Jul 16 11:01:56 Homebase kernel: __handle_mm_fault+0x881/0xcf9 Jul 16 11:01:56 Homebase kernel: ? raw_spin_rq_unlock_irq+0x5/0x10 Jul 16 11:01:56 Homebase kernel: handle_mm_fault+0x13d/0x20f Jul 16 11:01:56 Homebase kernel: do_user_addr_fault+0x2c3/0x48d Jul 16 11:01:56 Homebase kernel: exc_page_fault+0xfb/0x11d Jul 16 11:01:56 Homebase kernel: asm_exc_page_fault+0x22/0x30 Jul 16 11:01:56 Homebase kernel: RIP: 0033:0x14c2df9d424a Jul 16 11:01:56 Homebase kernel: Code: c5 fe 7f 07 c5 fe 7f 47 20 c5 fe 7f 47 40 c5 fe 7f 47 60 c5 f8 77 c3 66 0f 1f 84 00 00 00 00 00 40 0f b6 c6 48 89 d1 48 89 fa <f3> aa 48 89 d0 c5 f8 77 c3 66 66 2e 0f 1f 84 00 00 00 00 00 66 90 Jul 16 11:01:56 Homebase kernel: RSP: 002b:000014c2ce9ad8b8 EFLAGS: 00010206 Jul 16 11:01:56 Homebase kernel: RAX: 0000000000000000 RBX: 000014c2b408c800 RCX: 0000000000188c40 Jul 16 11:01:56 Homebase kernel: RDX: 000014c2b9119040 RSI: 0000000000000000 RDI: 000014c2b957f000 Jul 16 11:01:56 Homebase kernel: RBP: 00000000005eec00 R08: 0000000000000007 R09: 000014c2b408c820 Jul 16 11:01:56 Homebase kernel: R10: a539d7c265516a54 R11: 0000000000000090 R12: 0000000000000000 Jul 16 11:01:56 Homebase kernel: R13: 000014c2d4123908 R14: 0000000000000006 R15: 0000000000000000 Jul 16 11:01:56 Homebase kernel: </TASK> Jul 16 11:01:56 Homebase kernel: memory: usage 8388608kB, limit 8388608kB, failcnt 1505365 Jul 16 11:01:56 Homebase kernel: swap: usage 0kB, limit 9007199254740988kB, failcnt 0 Jul 16 11:01:56 Homebase kernel: Memory cgroup stats for /docker/e1ec82a8a67d06905519c219e24704d5e72978f69734455d13218e181a2a75e9: Jul 16 11:01:56 Homebase kernel: anon 4071133184 Jul 16 11:01:56 Homebase kernel: file 4457185280 Jul 16 11:01:56 Homebase kernel: kernel 61616128 Jul 16 11:01:56 Homebase kernel: kernel_stack 6701056 Jul 16 11:01:56 Homebase kernel: pagetables 18108416 Jul 16 11:01:56 Homebase kernel: sec_pagetables 0 Jul 16 11:01:56 Homebase kernel: percpu 4648 Jul 16 11:01:56 Homebase kernel: sock 0 Jul 16 11:01:56 Homebase kernel: vmalloc 8192 Jul 16 11:01:56 Homebase kernel: shmem 4457156608 Jul 16 11:01:56 Homebase kernel: file_mapped 101199872 Jul 16 11:01:56 Homebase kernel: file_dirty 0 Jul 16 11:01:56 Homebase kernel: file_writeback 0 Jul 16 11:01:56 Homebase kernel: swapcached 0 Jul 16 11:01:56 Homebase kernel: anon_thp 27262976 Jul 16 11:01:56 Homebase kernel: file_thp 0 Jul 16 11:01:56 Homebase kernel: shmem_thp 25165824 Jul 16 11:01:56 Homebase kernel: inactive_anon 4152025088 Jul 16 11:01:56 Homebase kernel: active_anon 20316160 Jul 16 11:01:56 Homebase kernel: inactive_file 24576 Jul 16 11:01:56 Homebase kernel: active_file 0 Jul 16 11:01:56 Homebase kernel: unevictable 4355948544 Jul 16 11:01:56 Homebase kernel: slab_reclaimable 28472680 Jul 16 11:01:56 Homebase kernel: slab_unreclaimable 8133008 Jul 16 11:01:56 Homebase kernel: slab 36605688 Jul 16 11:01:56 Homebase kernel: workingset_refault_anon 0 Jul 16 11:01:56 Homebase kernel: workingset_refault_file 418588 Jul 16 11:01:56 Homebase kernel: workingset_activate_anon 0 Jul 16 11:01:56 Homebase kernel: workingset_activate_file 37906 Jul 16 11:01:56 Homebase kernel: workingset_restore_anon 0 Jul 16 11:01:56 Homebase kernel: workingset_restore_file 11170 Jul 16 11:01:56 Homebase kernel: workingset_nodereclaim 0 Jul 16 11:01:56 Homebase kernel: pgscan 2603050 Jul 16 11:01:56 Homebase kernel: pgsteal 2224813 Jul 16 11:01:56 Homebase kernel: pgscan_kswapd 547864 Jul 16 11:01:56 Homebase kernel: pgscan_direct 2055186 Jul 16 11:01:56 Homebase kernel: pgsteal_kswapd 518175 Jul 16 11:01:56 Homebase kernel: pgsteal_direct 1706638 Jul 16 11:01:56 Homebase kernel: pgfault 13461787 Jul 16 11:01:56 Homebase kernel: pgmajfault 9769 Jul 16 11:01:56 Homebase kernel: pgrefill 286105 Jul 16 11:01:56 Homebase kernel: pgactivate 198791 Jul 16 11:01:56 Homebase kernel: pgdeactivate 213058 Jul 16 11:01:56 Homebase kernel: pglazyfree 0 Jul 16 11:01:56 Homebase kernel: pglazyfreed 0 Jul 16 11:01:56 Homebase kernel: Tasks state (memory values in pages): Jul 16 11:01:56 Homebase kernel: [ pid ] uid tgid total_vm rss pgtables_bytes swapents oom_score_adj name Jul 16 11:01:56 Homebase kernel: [ 24834] 0 24834 9410567 131038 2027520 0 0 jellyfin Jul 16 11:01:56 Homebase kernel: [ 4530] 0 4530 842285 111564 2748416 0 0 ffmpeg Jul 16 11:01:56 Homebase kernel: [ 7112] 0 7112 1023772 292197 4243456 0 0 ffmpeg Jul 16 11:01:56 Homebase kernel: [ 10540] 0 10540 668568 133634 1998848 0 0 ffmpeg Jul 16 11:01:56 Homebase kernel: [ 14498] 0 14498 711444 109988 2699264 0 0 ffmpeg Jul 16 11:01:56 Homebase kernel: [ 14718] 0 14718 770254 204448 3362816 0 0 ffmpeg Jul 16 11:01:56 Homebase kernel: [ 16731] 0 16731 221706 42837 1118208 0 0 ffmpeg Jul 16 11:01:56 Homebase kernel: oom-kill:constraint=CONSTRAINT_MEMCG,nodemask=(null),cpuset=e1ec82a8a67d06905519c219e24704d5e72978f69734455d13218e181a2a75e9,mems_allowed=0,oom_memcg=/docker/e1ec82a8a67d06905519c219e24704d5e72978f69734455d13218e181a2a75e9,task_memcg=/docker/e1ec82a8a67d06905519c219e24704d5e72978f69734455d13218e181a2a75e9,task=ffmpeg,pid=7112,uid=0 Jul 16 11:01:56 Homebase kernel: Memory cgroup out of memory: Killed process 7112 (ffmpeg) total-vm:4095088kB, anon-rss:1166848kB, file-rss:2284kB, shmem-rss:0kB, UID:0 pgtables:4144kB oom_score_adj:0 Jul 16 11:02:00 Homebase kernel: ffmpeg invoked oom-killer: gfp_mask=0x100cca(GFP_HIGHUSER_MOVABLE), order=0, oom_score_adj=0 Jul 16 11:02:00 Homebase kernel: CPU: 9 PID: 16883 Comm: ffmpeg Tainted: P O 6.1.74-Unraid #1 Jul 16 11:02:00 Homebase kernel: Hardware name: Gigabyte Technology Co., Ltd. B760M DS3H DDR4/B760M DS3H DDR4, BIOS F16 12/14/2023

On my test a few minutes ago the same log entry appeared right at the moment, when the 8GB memory limit of the Docker-Container was reached. The host systems memory had around 50% usage at that time.

I tried to reconstruct what happened on the 2nd of July and was successful. Back then I tested the encoding performance with multiple simultanious streams in Jellyfin. I could just recreate the OOM log entry. But what was running out of memory is the container which is limited to 8GB by me. Can you confirm that FCP also alerts me, when a container is running oom and not only the host system?

Thank you for your answer. I didn't configure it this way. I just tested it and it transcodes in the 'transcodes' folder unter appdata. I also limited the the amount of ram the container is allowed to use to 8GB.

Hello everyone, the fix common problems plugin just detected out of memory errors on my server. I'm not sure where to start digging... The only 'problem' that occurs quite regulary is that my jellyfin docker-container stops. In the log file of the unraid server the only 'out of memory' error entry I can find, seems to stop an ffmpeg process. The server has 32GB of Ram with about 8GB in use in 'normal' operation. Maybe you guys can give me a hint. Diagnostics are attached! Thank you very much in advance. Jaytie diagnostics-20240715-1910.zip

This helped, thank you! SourceMod is (for now) replaced with CounterStrikeSharp (CSS). And for CSS, admin-plugins are existing. There are all running with mySQL DBs which is a bit over the top for my usecase, but we will see

Thats what I also read. I will look into Sourcemod or the alternative for CS2 than, thx! Bit of a shame that something like that isn't integrated into CS dedicated server itself.

@ich777 Thanks for the CS2 Server Docker-Container, works great! Not 100% on topic, but do you guys have any recommendations on how to securely administrate the CS2 server remotely? There are a lot of mods out there like "fake rcon" an so on. This might work but is it safe enough? I'd be happy to hear, what you use.

Vielleicht noch eine kurze Rückmeldung von mir. Der Server läuft mittlerweile seit einigen Wochen. Bin recht zufrieden damit. Einziger Störfaktor ist derzeit noch das Gehäuse bzw. einige Eigenheiten des Gehäuses. Die 3,5" Aufnahmen passen nicht wirklich für Enterprise-Platten, da die Bohrungen an der Unterseite der Platten anders als bei Consumer HDDs sind. Kann das Gehäuse nicht wirklich was dafür und ist auch keine große Sache. Ingesamt gibt es noch einige Resonanzpunkte, wenn die HDDs laufen, die etwas nerven. Da muss ich nochmal ran. Ansonsten bin ich aber zufieden, zumal die Geräuschdämmung abseits der Resonanzen gut funktioniert. Der Rest der Hardware überzeugt auch. Der i5-12400 macht in meinem Anforderungsbereich echt Spaß. Der Boxed Kühler ist auch nach Stunden fast Volllast in dem Gehäuse nicht zu hören und die CPU pendelt sich dann bei max 62°C ein. Das Mainboard ist sparsam (danke an @DataCollector für den Tipp!). Der NVMe Kühler auf dem Mainboard ist top, die zweite NVMe wird mit einem günstigen Amazonkühler gekühlt, was auch gut funktioniert. Am Ende wurdes es übrigens eine 970 Evo Plus (Filecache) und eine 980 Pro (Anwendungscache).

Danke für die Erfahrungswerte. Tatsächlich hatte ich auch schon Probleme mit meiner SN570 500GB in einem BTFS Pool in Unraid. Seitdem der mit XFS formatiert ist, ist aber Ruhe. Naja drauf ankommen lassen will ichs nicht. Werden also vermutlich für den neuen Server dann Samsungs.

Das wär in meinem Sinn. Ich weiß nicht genau, was du mit "extrem intensiven Anwendungen meinst aber: Um die Lese- und Schreibgeschwindigkeit geht es mir hierbei nicht, da ich die mit meinem Netzwerk (wie du auch schreibst) nicht ausreize. Mir gehts eher um die Leistung der SSD als Anwednungsspeicher: Weder noch, eher ein NAS + Docker- und VM Host. Auf einer der beiden SSDs werden halt Dockeranwedungsdaten liegen. Die potentiell anspruchsvollsten werden 1-2 Gameserver (jeweils 10 Slots), Jellyfin mit max zwei Transcodes, Synapse Chatserver und Nexcloud mit einigen Usern und deren Datenbanken sein). Dazu kommen noch 1-2 virtuelle Maschinen. Windows und Ubuntu. Einen Teil davon betreibe ich schon jetzt auf einer WD Blue SN570, welche ja auf dem Papier eine schlechtere Performance mitbringt. Möchte den neuen Speicher nur nicht direkt an die Grenze treiben, wenn jetzt noch Gameserver und Jellyfin hinzu kommt. Gerade weil die Preise zwischen einer 970 Evo Plus und einer 980/990 Pro oder SN770/Sn850X sehr ähnlich sind. Denke aber ich werd auf die 970 Evo Plus setzen.

Danke! Dh. du würdest noch auf PCIe 3.0 zugunsten der Kompatibilität setzen?

Das Nanoxia Deep Silence 4 hab ich mittlerweile hier und bin recht positiv überrascht und recht zufrieden damit. Der oberste und unterste HDD Einschub hat einen Konstruktionsfehler, da die Gewinde die den Käfig befestigen an den Platten schleifen, wenn man sie rein schiebt. Mal schauen ob ich das noch glatt schleife. Gibts Erfahrungswerte, was NVMEs angeht? Man liest immer mal wieder was von Kompatiblitätsproblemen mit Unraid von diversen Speichern. Mal mit WD SN770/750 aber auch mal mit Samsungs. Es ist aber nie wirklich klar, ob es wirklich an den Flashspeichern lag... Ich suche 2x1TB für den Schreibcache und für Docker + VMs. Preise sind derzeit ja eher hoch. Aber ins Auge gefasst hab ich SN770/SN850X bzw. Samsung 980Pro/990Pro. Könnt ihr was empfehlen?