[BUG] JELLYFIN_CONFIG_DIR=/config makes .NET watch the entire /config tree (incl. metadata), exhausting inotify watches

Đang mở
#341 2 bình luận 0 reaction 0 người được giao Xem trên GitHub

Chưa có ai nhận issue này.

Đánh giá

Độ khó
4/5
Thời gian dự kiến
3-5 ngày
Mức phù hợp với người mới
25/100
Loại issue
Lỗi
Độ rõ ràng
Khá rõ ràng
Mức độ hoạt động
Sôi nổi
Công nghệ
csharp, docker, dockerfile, linux
Lĩnh vực
devops, infrastructure

Hướng nghiên cứu

Start with the svc-jellyfin run script, where JELLYFIN_CONFIG_DIR is exported as /config, and with the Dockerfile that sets the default config path. Confirm the watch count by comparing inotify watches to the directories under /config. Done means Jellyfin only watches its config files, with existing installs migrated safely, and the choice between a /config/config subfolder and a Dockerfile-level override is settled first.

Do mô hình lập chỉ mục viết ra từ nội dung của issue.

Mô tả

no-issue-activity
Is there an existing issue for this?
  • I have searched the existing issues

Related: #331 (same symptom, attributed to real-time monitoring) and #22 (config path differs from the official image).

Current Behavior

The svc-jellyfin run script exports JELLYFIN_CONFIG_DIR="/config". Jellyfin registers its logging config with SetBasePath(appPaths.ConfigurationDirectoryPath) and AddJsonFile(..., reloadOnChange: true) (Jellyfin.Server/Program.cs ~L400-403). On Linux, .NET's physical file provider watches that base path recursively (one inotify watch per subdirectory). With the config dir set to /config, that includes /config/data/metadata, /config/cache, etc.

Measured on a host running linuxserver/jellyfin 12.2:

  • /config/data/metadata holds ~96,000 directories (mostly People/*).
  • The Jellyfin process holds 117,465 inotify watches, all on the /config filesystem (sampled inodes resolve to /config and /config/data/metadata/People/...).
  • Disabling real-time monitoring on every library and restarting changes nothing (still 117,465).
  • Because the limit is per UID, other containers running as the same PUID run out of watches. In our case Plex's EasyAudioEncoder failed with EAE timeout! EAE not running, or wrong folder?.

This is why #331 saw exhaustion that did not go away with real-time monitoring settings.

Expected Behavior

Only Jellyfin's actual config files (system.xml, logging*.json, ...) are watched, as with the official image, where the config dir is a dedicated subfolder and the watch count stays small.

Possible backward-compatible fix: point JELLYFIN_CONFIG_DIR at a dedicated folder (e.g. /config/config) with a one-time init step that moves *.xml / logging*.json there if they still sit in /config. Or, as suggested in #22, move the env into the Dockerfile so users can override it.

Steps To Reproduce
  1. Run linuxserver/jellyfin with a library large enough to produce a few thousand metadata folders.
  2. Count watches held by Jellyfin: cat /proc/$(pgrep -f /usr/bin/jellyfin)/fdinfo/* | grep -c '^inotify wd:'
  3. Compare with find /config -xdev -type d | wc -l inside the container: the watch count tracks the total number of directories under /config.
  4. Disable real-time monitoring for all libraries and restart: the count does not change.
Environment
  • OS: Debian-based umbrelOS 2.0 (host 1); Ubuntu 24.04 VPS (host 2, small library, same pattern: ~2 watches per directory under /config)
  • How docker service was installed: distro packages / umbrelOS
CPU architecture

x86-64

Docker creation
services:
  jellyfin:
    image: linuxserver/jellyfin:version-12.2ubu2604
    environment:
      - PUID=1000
      - PGID=1000
    volumes:
      - ./jellyfin:/config
Container logs
No Jellyfin errors; the effect shows up in other processes under the same UID, e.g. Plex:
[eac3_eae] EAE timeout! EAE not running, or wrong folder? Could not read '/run/plex-temp/.../EasyAudioEncoder/Convert to WAV (to 8ch or less)/...'

Host: fs.inotify.max_user_watches = 122404; watches in use by UID 1000 = 122386 (Jellyfin 117465, Syncthing 4751)

Investigated and written with the help of an AI assistant; all numbers above were measured on my own servers.

Ngôn ngữ chính
Dockerfile
Star
909
Fork
136
Merge trung bình
4 giờ 31 phút
Pull request đã merge (30 ngày)
1

Chuẩn bị môi trường

Bắt đầu từ đâu

  1. Đọc hết issue, rồi đọc hướng dẫn đóng góp của dự án.
  2. Bình luận trên issue rằng bạn sẽ nhận — tránh hai người làm cùng một việc.
  3. Fork repository và làm thay đổi trên một nhánh.
  4. Mở pull request có tham chiếu số hiệu của issue.

Issue khác của linuxserver/docker-jellyfin

Tất cả issue của linuxserver/docker-jellyfin

Issue tương tự

Thêm issue về DevOps

Nhận issue mới trong hộp thư của bạn

Bản tóm tắt ngắn những issue GitHub phù hợp với người mới.