Error: EACCES: permission denied, mkdir '/.ffbinaries-cache' on start

[CarbonCollins]

Hi,

Just tried to load the docker image into a helm chart to start using it, however, when I first started the image it crashed with the following error:

node:fs:1325
  const result = binding.mkdir(
                         ^

Error: EACCES: permission denied, mkdir '/.ffbinaries-cache'
    at Object.mkdirSync (node:fs:1325:26)
    at Kre (/fp/float.cjs:105:3995)
    at /fp/float.cjs:105:4009
    at /fp/float.cjs:1:263
    at /fp/float.cjs:105:8659
    at /fp/float.cjs:1:263
    at Object.<anonymous> (/fp/float.cjs:298:83474)
    at Module._compile (node:internal/modules/cjs/loader:1760:14)
    at Object..js (node:internal/modules/cjs/loader:1892:10)
    at Module.load (node:internal/modules/cjs/loader:1480:32) {
  errno: -13,
  code: 'EACCES',
  syscall: 'mkdir',
  path: '/.ffbinaries-cache'
}  

I was using the latest image which resolved to this image at the time: https://hub.docker.com/layers/inrix/floatplane-downloader/latest/images/sha256-ebc073d0cc20abc7ac77d8a9e22f7d94674dfa3dd02cfc3a3a24ea7ab250b4c6.

I'm not sure how to solve this as I searched previous issues and the wiki which didnt really make any mention to this specific issue nor does there seem to be any way of specifying what UID and GID is used in the container

[Inrixia]

Are you running on unraid? This is a permissions issue with your host filesystem

[CarbonCollins]

No i’m running in a kubernetes cluster with the /fp/videos mounted to an nfs share and /fp/db mounted to an iscsi volume created by the cluster for the container.

The path it’s complaining about in the stack trace is within the container but not within one of the volume mounts so I don’t think it’s the volumes in this case.

[Inrixia]

Hmm, this is definitely some permissions issue with the container on k8s.

Nothing I can do, you need to figure out why it's not allowed to write to that file.

I expect maybe k8s has stricter perms for that but idk why that would be the case...

[Inrixia]

You could try doing a ephemeral empty volume mount for that folder, it'd likely fix it.

[bunnypranav]

@Inrixia I am getting a similar error, any suggestions on what I can do?

[12/20/2025, 8:54:51 AM] Searching for new videos in LTT Supporter
[12/20/2025, 8:54:55 AM] Queued 1 videos...
[12/20/2025, 8:54:55 AM] Luke Week 2: How a New Studio Made the Multiplayer Game of the Year - Saving captions
[12/20/2025, 8:54:55 AM] Luke Week 2: How a New Studio Made the Multiplayer Game of the Year - Saved captions
[12/20/2025, 8:54:55 AM] Luke Week 2: How a New Studio Made the Multiplayer Game of the Year - Muxing ffmpeg metadata...
[12/20/2025, 8:54:55 AM] Luke Week 2: How a New Studio Made the Multiplayer Game of the Year - An error occoured: Muxing ffmpeg metadata - spawn ./db/ffmpeg EACCES

For testing purposes, I did chmod 777 -R to the db and videos directories, but still no use. I have mounted that folder from an ntfs hard drive on Ubuntu 24.04 if it helps.

[Inrixia]

That ffmpeg binary needs to be executable, so you could try with chmod +x

[bunnypranav]

Did that, but did not change anything.

[Inrixia]

There is a pinned comment in discord that might help see https://discord.com/channels/468977142198435841/468977376794247199/1199142897593892974

Otherwise I can't help much, its a filesystem permissions error with your mount.

[CarbonCollins]

going back to the binaries cache issue, I slapped a bind mount onto /.ffbinaries-cache and still had my securityContext set on the pod (setting the user and group that the container should run as) which seems to run now as I could log in and it began downloading episodes and art work. Im wondering why the rest of the container is able to operate correctly but just the creation of the binaries cache fails...

[Inrixia]

Well if you have you'd db in a mount the container doesn't try write anywhere else. But it is weird irs getting iffy about that. Probably a k8s thing

[CarbonCollins]

Yea the db is mounted into a persistant volume.
I have 3 volumes mounted now at:

/.ffbinaries-cache -> volitile iSCSI mount
/fp/db -> persistant iSCSI mount
/fp/videos - > NFS mount

I tried to look into how the unraid template worked but I dont know enough about how unraid works to find the actual template and how that works to compare :P only that it has the option to specify the PUID and GUID somehow. If thats mapping into an ENV var somewhere that could be the issue that I was having as I do not specify any PUID or GUID env vars to tell the container itself.