This is the support thread for multiple Plugins like:

 

- Cycle VM

## 20241208 updated Docker feature

## 20241207 initial release

 

- Pwm Fan

## 2025.01.21 - crond email push fix

## 2024.12.31 - added a loop lookup for correct temp, compare and take lower value, thanks @avargaskun for the catch

## 2024.12.21a - added aggressiv mode for CPU mode (more demanding flexible CPU cooling)

## 2024.12.21 - added extra HDD Section for split setup (following main temp fan rules)

## 2024.12.18 - fixed log fetching (less spam)

## 20241217 - autostart typo fixed in routine

## please disable / enable the selected autostarts after update to apply to the existing autostarts.

## 20241214 - cosmetics and added log page

## 20241207 - initial release

 

- Cache Mover (Media Caching)

 

## 2025.09.06 - python existing lookup - is python plugin is installed, skip install / uninstall

## 2025.02.18 - small emby / jelly return fix for special chars

## 2025.02.17 - fixed typos to reduce spam

## 2025.02.16 - added http / https Mediaserver lookup - sorting pushbits Notifications - hopefully finally sorted special Chars in folder / filenames

## 2025.02.03 - may bug fix - jq errors - other fixes

## 2025.01.27 - cache mover additional options besides regular cache - array setup - see release notes

## 2025.01.21 - crond email push fix

## 2025.01.12a - fixing brackets in mover cleaning list thanks @simatra - cleaning fix for empty exclusion entries.

## 2025.01.11   - included a MediaServer Option - ONLY with /mnt <> /mnt setup in Mediaservers !!! update will follow to rewrite - Update done, no more limitation
## 2024.12.25 - included a exclusion Option - more aggressiv watching for write files to skip execution

## 2024.12.23 - added file write check and break processing

## 2024.12.22 - fixed 6.12.13+ compatibility and updated gdb and python on v7

## 2024.12.18 - fixed log fetching (less spam)

## 20241217 initial beta release, feedback welcome, be aware about settings, this plugin could delete files if missconfigured !!!

 

- ASPM Helper

## 2024.12.18 fixed log fetching (less spam)

## 20241214 cosmetics and added log page

## 20241211 initial release

 

- Simple Mover

## 2025.02.22 - small bug fixes - option to turn off standard mover to run as only mover option

## 2025.01.21 - crond email push fix

## 20250111 - initial release beta

 

- FFmpeg Tool

## 202501nn to follow

---

open todo

cache mover

- check free space on target pool (adjustable) >> done and testing here - done, testing here

- handle unbalanced files fetching (mainly tv shows) / together with multi array support - open

- multi Media Server support (plex, emby, jelly simultan) - done, testing here

- multi KeepHours support (handle sep TV, Anime, Movie, ...) - open

- zfs issues ? (couldbt be replicated here) - open

- better UI (dropdown) - open

pwm fan

- keep min rpm level when disk is off (as optional choice) >> done and testing here - done, testing here

- add extra hdd block (having 4x hdd blocks, 1x single cpu) >> done and testing here - done, testing here

- better UI (dropdown) - open

 

Descriptions:

 

- Cycle VM

 

sample usage, 2 Desktop VM/s sharing the same dGPU, while sitting on the Mashine running Desktop Windows and want to switch to lets say Linux without turning Windows VM off, without taking a separate client turn Linux VM on, this will

- stop the Current running VM

- wait until its shutdown

- start the other VM

- cycle docker <> vm in terms dGPU is used by a docker

- - BEWARE, only if the dGPU is NOT vfio bind dependant

 

- Pwm Fan

 

this plugin is intended for low Power systems which are most liekely passive cooled, but as disks and cpu in usage likes cooling and also the CPU should be considered in usage, this plugin has the following features

- check CPU Temp and control 1 Fan (CPU Fan) therefore

- check hdd disk temps and take highest value as control multiply fans therefore

- currently only supported by cron, so 1 min is lowest level for lookups !!!

- WARNING, CPU Fan is always recommended to be controlled by BIOS and not Software, no warranty here.

 

- Cache Mover (Media Cacher)

this plugin will be a read cache for Media lying on the array disks, so this plugin will work as follows

- check array activity every N minutes

- if a media (mkv, mp4, ts, ...) is in usage for more then N seconds

- copy (recommended) or move it to a pool (ssd)

- - split handling here, movies single, tv shows always the full season.

- switch internally to pool (ssd) path so the array disk can go back to sleep

- - 2 modes possible, easy (cut) file access, Player will stop, restart Stream will prefer pool, hard (rewrite internally) switch to pool "live" (switch working on v7 only, fixed with 20241223)

- - prerequest here, Media Servers are accessing /mnt/user/ and not /mnt/disk1, /mnt/cache/, /mnt/... like Plex, Emby, ... same for smb

- keep file/s on pool (ssd) for N hours, split setup (2 choices like TV Shows -longer- Season or Movies -shorter-)

- - only working when the mover is not moving daily, hourly, ... files from this pool

- - remove files (or move back) after the given last access time (2 choices), if they are copied, you can either wipe or move to recycle bin if used.

- mover tuning ignore list integration (lists will always be merged)

- - BEWARE, while unbalancing disable this plugin !!! (move files from diskN > diskN in array)

- - OR USE MediaServer Option ONLY

 

- ASPM Helper

this plugin will help finding the correct way to force some devices in the correct aspm state for better power consumption and set it on autostart too.

- always to consider, forcing aspm can result in instability, undesirable behavior and so on ...

- will advise which devices are currently not in a decent aspm state

- setup then 3 points

- endpoint, root complex, aspm level mode and find the correct command and set it.

- experimental ... working fine here on 4 mashines, but also already had some trouble with this feature

 

- Simple Mover

simple mover addon, mainly considered for Media Files to keep on cache
NO replacement for regular mover or mover tuning, actually mover tuning is mandatory for exclusion
2 options

advanced, shares/subfolders moving by rules, age based, size based, treehold when > to % usage

basic, simple move, only option min age for basic Media moves

both can be triggered seperately (cron)

 

- FFMpeg Tool

this plugin is "abusing" a working docker which has hw accel ffmpeg running, most likeley emby as sample. as im encoding also live tv recordings which are interlaced and none of the Media Converters fitted to me, i wrote me something up to also compare and so on.

some FFMpeg affinity is mandatory here.

- will convert existing Media by choices (advanced)

- vaapi, qsv, nvenc supported

- source path, temp path, replace the original, keep a backup for x days in some backup place for original files, and so on

- fully custom ffmpeg commands which also

- - run then the ffmpeg by your command and

- - or auto crop black bars

- - deinterlace for real (sample real 50i live broadcasts) or encode interlaced progressiv ()sample fake 50i movies, shows, ...) depending on source

- - and so on

Cycle VM: Could this be extended to cycle docker containers that use the GPU? For example, I usually have an AI running that uses GPU, but if I forget to turn it off and I start a VM that also uses the GPU, system lockup.

1 hour ago, The_Holocron said:

Cycle VM: Could this be extended to cycle docker containers that use the GPU? For example, I usually have an AI running that uses GPU, but if I forget to turn it off and I start a VM that also uses the GPU, system lockup.

ok, never thought of that (yet)

 

noted, and yes, no deal breaker ... just have to think about how to tag either Docker or VM ... will leave a note here.

 

thanks for the feedback.

14 hours ago, The_Holocron said:

Cycle VM: Could this be extended to cycle docker containers that use the GPU? For example, I usually have an AI running that uses GPU, but if I forget to turn it off and I start a VM that also uses the GPU, system lockup.

 

this will be added in the next release, so 2 ways, either VM <> VM or VM <> Docker/s

 

Hi the fan plugin seem not working need additionnal thing that are not visible ?

 

2 hours ago, francrouge said:

Hi the fan plugin seem not working need additionnal thing that are not visible ?

 

shouldnt be, what errors are you getting ?

Im trying to use cycle vm plugin , i have put the name of VM and the names of two docker containers from whom i want to release the dgpu , but when i click cycle docker and head over to the vm tab my unraid gui freezes and no vm can be seen . Only way to get out of this situation is a hard reboot . I feel the cycle vm docker is not working to switch dgpu between docker containers and vm. The dgpu is not getting released to give it to the vm 

Edited December 10, 20241 yr by drmetro

38 minutes ago, drmetro said:

I feel the cycle vm docker is not working to switch dgpu between docker containers and vm

i ll look into it, thanks for the feedback so far.

 

may please post a screen from your setup page and the flow where it freezed was

 

2 dockers where running using the dGPU, and Cycle did freeze your Server due not stopping dockers before starting the VM, correct ?

6 minutes ago, alturismo said:

i ll look into it, thanks for the feedback so far.

 

may please post a screen from your setup page and the flow where it freezed was

 

2 dockers where running using the dGPU, and Cycle did freeze your Server due not stopping dockers before starting the VM, correct ?

Yes , i tried two times after hard reboot( power down ) , both times webui freezes on clicking the VM tab , basically i feel whats happening is that the dGPU is not getting released from docker and passed to the VM so only its freezing, before i used to manually free the gpu and pass it to the VM from system devices followed by a reboot but after seeing this plugin i had high hopes that single click of button can do it but its not working.

currently the webui is not opening and i cant hard reboot because mover is running so I cant post any pictures, the logs also didn’t show anything significant or error.

Edited December 10, 20241 yr by drmetro

6 minutes ago, drmetro said:

before i used to manually free the gpu and pass it to the VM from system devices followed by a reboot but after seeing this plugin i had high hopes that single click of button can do it but its not working.

ahh, ok, so manually it also doesnt work in your case ?

 

like, the dGPU is locked on boot by the OS and only vfio bind (and reboot) will make it usable for your VM ?

Just now, alturismo said:

ahh, ok, so manually it also doesnt work in your case ?

 

like, the dGPU is locked on boot by the OS and only vfio bind (and reboot) will make it usable for your VM ?

Yes , only unbinding than reboot makes the gpu free than I attach it to VM and than it works for VM only and cannot be used for docker containers . To make it work for docker container i have to bind it , rebbot ,  than it works for dockers and now cannot be used for Vm. I was hoping this plugin will make the binding /unbinding smooth without need for reboot which will solve this problem.

1 minute ago, drmetro said:

I was hoping this plugin will make the binding /unbinding smooth without need for reboot which will solve this problem.

sorry, there is currently no way to "live" vfio bind / unbind a dGPU, now its clear, thanks for the Info

 

its more meant to be used if the dGPU can be used in a VM also without vfio bind ticked ...

 

sample here, intel iGPU is primary (and used by host Unraid), 2x Nvidia Cards are here which can also be used either / or on Host (Dockers) or in a VM.

 

i make a not, thanks for reporting and bringing this up

1 minute ago, alturismo said:

So , this plugin wont be of any help ?

1 minute ago, drmetro said:

So , this plugin wont be of any help ?

not in this case, sorry ... vfio bind dependency is a "deal breaker" ... there is no way around if your sys needs this behaviour

If vfio bind is not ticked , the dGPU cannot be used for dockers … it can be only used for VM . So the switching vm to docker in the plugin doesn’t serve any purpose. Only vm to vm will be helpful.

7 minutes ago, drmetro said:

If vfio bind is not ticked

i guess you mean "ticked" ...

 

if you tick vfio bind, a device is "invisible" for the host, so dockers cant use it ... only be passed to a VM then.

 

if its untcked AND not in general use by the Host, sample Unraid for itself, then it can be EITHER used by Dockers OR passed to a VM, in this case this plugin will do its job

 

Cycle ... look if the VM is running, if yes, stop the VM, wait till its shutdown (dGPU is free), start the docker/s ... if the VM is not running, stop the docker/s, wait till they stopped (dGPU is free), start the VM ...

 

f its untcked AND in general use by the Host, sample Unraid for itself, then it can only be used for docker/s ...

1 minute ago, alturismo said:

if its untcked AND not in general use by the Host, sample Unraid for itself, then it can be EITHER used by Dockers OR passed to a VM, in this case this plugin will do its job

but if its unticked dockers cannot use the dGPU it doesn’t even show on gpu statistics on dashboard, can you please explain this 

Just now, drmetro said:

but if its unticked dockers cannot use the dGPU it doesn’t even show on gpu statistics on dashboard, can you please explain this 

may post a screen from your devices when you can access your server (dGPU part), then i rather can better explain it

So after unticking vfio bind , this plugin can effectively give the dGPU to vm and docker ? thats what i want .. but after i untick the docker containers cannot see the gpu how can this plugin give the docker container the dGPU

1 minute ago, alturismo said:

may post a screen from your devices when you can access your server (dGPU part), then i rather can better explain it

Yes i will , now tye webui is not accessible i have to hard reboot but im waiting for mover to finish after that i will hard reboot 

Just now, drmetro said:

So after unticking vfio bind , this plugin can effectively give the dGPU to vm and docker ?

only if your sys can handle this ... so you can test it simple, start your server without vfio binds like this (empty not ticked dGPU)

 

 

then post a screenshot here, it should be empty like this sample while the VM and Docker/s are not using it.

 

then simple start your VM, if it runs like this and check again, the field should be grey (in use) ... if you turn VM off again, it should be free again ... and free to use for another VM, or Docker/s, or ... but only if its NOT vfio bindes in any way.

 

may also post a diagnostics to make sure there is no vfio binds setted ...

If its unticked only VM can use it , the docker cannot use the dgpu because they cannot see the dgpu even if vm is not using the gpu , the vfio unbound gpu can be used only by VMs

Edited December 10, 20241 yr by drmetro

1 minute ago, drmetro said:

If its unticked only VM can use it , the docker containers because they cannot see the dgpu even if vm is not using the gpu

let us take a look when you can access your server ... as you describe it "vice vers" the expected behaviour ...

 

ticked (vfio bind) == ONLY passed usage for a VM

unticked (not vfio bind) == free usage (if possible and not grabbed by the OS)

 

more complex sample here

 

in this constellation i can use the Nvidia Cards either / or ...

 

they are NOT vfio bound as you see, they are currently NOT in use by anything, which usually ONLY works when you also have a iGPU, as you see my 1st device is a Intel iGPU and is set as primary device (BIOS) ... so the Nvidias are NOT used by Unraid by default ... 

 

 

but, if i would now start a VM and simultan also a Docker using the same dGPU ... the system will most likely crash ... therefore this plugin is intended to turn 1 side off before the other is using it ...

 

if i know would do this and reboot

 

 

the Nvidia GTX1060 could ONLY be used in a VM and definately not in a docker anymore as its "bind to vfio ONLY" as a passthrough device

 

thats how its supposed to be, so lets look how your setup looks like and we see further

I don’t have any igpu , on my setup

unbind vfio = makes dgpu unavailable for unraid , dockers even nvidia plugin doesn’t recognise it because gpu is detached from unraid , its free and only can be passthrough for VM and only VM.

bind vfio = the dgpu becomes available for unraid and dockers and can be used by unraid and dockers and not by VM . If i start vm it all crashes .