pabloromeo
commented
1 year ago Hi!
So, given that you are seeing content in the /transcode directory getting generated, I assume it's mostly configured correctly.
Can't say for sure regarding the NVIDIA settings, as I don't have an environment to replicate it. But I've seen other users using GPUs in their configurations.
Now, one thing I noticed is that you have different PUID and PGID between Plex and the worker. Could it be that you are having permissions issues when Plex is trying to read those files? I assume they are getting created and owned by UID 1000, so Plex might not be able to read them.
You could try changing permissions on /volume1/docker/plex/transcode to 777 just to see if Plex is able to pick up the content.
The reason for the playback not starting could be multiple, to name a few that I remember:
- Plex not being able to read the files in
/transcodedirectory due to permissions - Plex not receiving the "progress" http calls coming from the worker through the local relay
- A problem with the shared storage used between Plex and the workers (NFS? SMB? something else?)
Also you might find relevant info in Plex's own logs, which are available in their UI, under Settings -> Manage -> Console.
Is this on docker swarm or just standalone docker on each of those two?
paytah232
The transcode did seem to fire up better after that but it still seems really rather slow. Sometimes it plays for a few seconds and stop, other times, I only get the first frame preview or a total failure. Eventually, the stream shuts itself down.
github-actions[bot]
Hi @pabloromeo, I've been trying my hardest to get this system working smoothly and I think I'm out of luck,so I'm hoping to get your assistance. I'm not sure exactly what you will require from me, but I will get whatever logs / files you need.
My setup:
services: plex: container_name: plex image: ghcr.io/linuxserver/plex:latest environment: DOCKER_MODS: "ghcr.io/pabloromeo/clusterplex_dockermod:latest" VERSION: docker PUID: 1027 PGID: 65538 TZ: Australia/Brisbane ORCHESTRATOR_URL: http://plex-orchestrator:3500 PMS_SERVICE: 192.168.0.2 # This service. If you disable Local Relay then you must use PMS_IP instead PMS_PORT: "32400" TRANSCODE_OPERATING_MODE: both #(local|remote|both) TRANSCODER_VERBOSE: "1" # 1=verbose, 0=silent LOCAL_RELAY_ENABLED: "1" LOCAL_RELAY_PORT: "32499" ADVERTISE_IP: http://192.168.0.2:32400/ healthcheck: test: curl -fsS http://localhost:32400/identity > /dev/null || exit 1 interval: 15s timeout: 15s retries: 5 start_period: 30s volumes:
32414:32414/udp restart: always
plex-orchestrator: container_name: plex_orchestrator image: ghcr.io/pabloromeo/clusterplex_orchestrator:latest healthcheck: test: curl -fsS http://plex-orchestrator:3500/health > /dev/null || exit 1 interval: 15s timeout: 15s retries: 5 start_period: 30s environment: TZ: Australia/Brisbane LISTENING_PORT: 3500 WORKER_SELECTION_STRATEGY: "LOAD_RANK" # RR | LOAD_CPU | LOAD_TASKS | LOAD_RANK (default) volumes:
The following is my worker config, which is running on an old laptop with an NVIDIA GPU,with the drivers installed and reporting CUDA working etc. `version: '3.8'
services: plex-worker: image: ghcr.io/linuxserver/plex:latest hostname: "plex-worker" container_name: plex_worker deploy: mode: replicated replicas: 2 devices:
runtime: nvidia`
Using these, I am able to get the workers talking to the orchestrator, however, it appears that the workers do increase the GPU usage by about 200MB, but the CPU is still massively high (strangely reporting 195% usage). If I access the transcode directory,files are appearing, and it seems the size of them should be sufficient for the playback to start, but it is just continuously loading, until it eventually fails.
I also noted like in a previous issue, that the plex log does seem to report the request coming from a non-local IP (my external IP), even though everything is local. Not sure if that is the issue, but if you are able to help me troubleshoot why the setup isn't working for me, I'm hoping it might help others in similar situations fix it too.
Thanks!