Transcoding failure: EAE watchfolder is not writable

[craigcabrey]

Describe the bug

Intermittently seeing errors of EAE watchfolder is not writable in the worker logs. I'm not able to reliably reproduce it. Seems to be a potential re-occurrence of #244.

Additional context

[Parsed_scale_0 @ 0x7f717868a5c0] w:2560 h:1280 flags:'' interl:0
[eac3_eae @ 0x7f71784acb80] EAE watchfolder is not writable: /tmp/pms-01601f0f-6718-407c-8327-356a1a732cb7/EasyAudioEncoder/Convert to WAV (to 8ch or less)/jzddbmz6b5hy352mv8xod9e2_977-0-test.tmp
Stream mapping:
  Stream #0:0 (hevc) -> scale:default (graph 0)
  Stream #0:1 (eac3_eae) -> aresample:default (graph 1)
  format:default (graph 0) -> Stream #0:0 (libx264)
  aresample:default (graph 1) -> Stream #0:1 (aac)
Error while opening decoder for input stream #0:1 : Generic error in an external library
[AVIOContext @ 0x7f717fd48700] Statistics: 77498 bytes read, 2 seeks
Transcoder exit: child process exited with code 1
Completed transcode
Removing process from taskMap
Transcoder close: child process exited with code 1

The error is misleading though, since that user can write to tmp:

root@prod-plex-worker-1:/# su - abc -s /bin/bash
abc@prod-plex-worker-1:~$ cd /tmp/pms-01601f0f-6718-407c-8327-356a1a732cb7/EasyAudioEncoder
abc@prod-plex-worker-1:/tmp/pms-01601f0f-6718-407c-8327-356a1a732cb7/EasyAudioEncoder$ touch test
abc@prod-plex-worker-1:/tmp/pms-01601f0f-6718-407c-8327-356a1a732cb7/EasyAudioEncoder$ ls -l
total 0
-rw-rw-r--. 1 abc abc 0 Sep  1 18:03 test
abc@prod-plex-worker-1:/tmp/pms-01601f0f-6718-407c-8327-356a1a732cb7/EasyAudioEncoder$

[Nullvoid3771]

Make sure config folders are set to permission 0777. Avoid using SMB for config, transcode, codecs. Use cephs if possible.

See #335

[craigcabrey]

I am using CephFS.

[Nullvoid3771]

@craigcabrey The fix for this is to create multiple temp folders in cepfs. Then create multiple workers with different names. constrict each worker and pms to a single replica and locked to specific nodes. Make sure your tmp folders have write permissions. Adapt to your enviroment and add appropriate volume mappings. You don't even need to do this in ceph you can do it in local volumes, because you’re constraining your containers to a specific node. This sounds counter intuitive since you’re either using K8 or swarm, but it's what I found worked. If you share tmp folders you will run into conflicts.

Docker Swarm Stack - Sry I don’t use K8

services:
 plex:
   image: ghcr.io/linuxserver/plex:latest
   deploy:
     placement:
       constraints:
         - node.hostname == node1
     mode: replicated
     replicas: 1

volumes: 
- type: bind             
        source: $HOST/mnt/pve/cephfs/tmp
        target: /tmp 


plex-worker:
   image: ghcr.io/linuxserver/plex:latest
   hostname: "plex-worker-{{.Node.Hostname}}"
   deploy:
     placement:
       constraints:
         - node.hostname == node2
     mode: replicated
     replicas: 1

volumes: 
- type: bind             
        source: $HOST/mnt/pve/cephfs/tmp2
        target: /tmp 


plex-worker2:
   image: ghcr.io/linuxserver/plex:latest
   hostname: "plex-worker2-{{.Node.Hostname}}"
   deploy:
     placement:
       constraints:
         - node.hostname == node3
     mode: replicated
     replicas: 1

volumes: 
- type: bind             
        source: $HOST/mnt/pve/cephfs/tmp3
        target: /tmp 

[pabloromeo]

IIRC /tmp folders don't actually require any sharing at all, they don't even need specific bind mounts.
What needs to be shared across workers and PMS is the transcode location (and cache and drivers for hardware transcoding to work).