Skip to content

Tdarr

Tdarr is an automated media transcoding application. It scans media libraries and converts files to a target codec (H.265/HEVC), reducing storage usage while maintaining quality. It uses Intel QuickSync for hardware-accelerated transcoding.

Details

Property Value
Helm chart app-template v4.6.2 (bjw-s)
Image ghcr.io/haveagitgat/tdarr:2.65.01
Ports 8265 (web UI), 8266 (server)
HTTPRoute tdarr.homelab.local
Namespace arr
ArgoCD app arr-tdarr
Sync wave 1
Internal URL http://arr-tdarr.arr.svc.cluster.local:8265

Storage

Volume Type Size Mount Path Notes
config PVC (nfs-client) 1Gi /app/server Tdarr server database and configuration
temp PVC (nfs-client) 100Gi /temp Scratch space for active transcodes
media PVC (existing arr-data) -- /data/media Shared media library (subPath: media)
dri hostPath -- /dev/dri Intel GPU device for hardware transcoding

Resources

CPU Memory
Requests 200m 512Mi
Limits -- 4Gi

GPU limit: gpu.intel.com/i915: 1

Scheduling

  • nodeSelector: gpu: intel

Key Configuration

  • Environment variables from ConfigMap arr-env (TZ, PUID, PGID) plus application-specific variables:

    Variable Value
    serverIP 0.0.0.0
    serverPort 8266
    webUIPort 8265
    internalNode true
    inContainer true
    ffmpegVersion 6

  • Runs with an internal processing node (internalNode=true), so no external Tdarr node deployment is required.

  • Ingress annotations set proxy-read-timeout and proxy-send-timeout to 3600 seconds to prevent nginx from timing out during long transcode operations.
  • Liveness, readiness, and startup probes check the web UI on port 8265. The startup probe allows up to 30 failures at 10-second intervals (5 minutes).

Post-Deploy Setup

1. Verify the Internal Node

  1. Open https://tdarr.homelab.local.
  2. Go to the Tdarr tab (main dashboard). The internal node "InternalNode" should appear automatically.
  3. Click the node name, then Options. Set workers:
    • Transcode CPU workers: 1 (remuxing is I/O-bound; more workers on NFS won't help)
    • Transcode GPU workers: 0 (not needed for remux/copy operations)
    • Health check CPU workers: 1
    • Health check GPU workers: 0

2. Create the Remux Flow

Go to the Flows tab and click Flow+ to create a new flow. Build this pipeline using the visual node editor:

  1. Input (auto-created)
  2. Check File Medium (category: file) -- ensures it's a video file
  3. Begin Command (category: ffmpegCommand) -- starts building the ffmpeg command
  4. Set Container (category: ffmpegCommand) -- set to mkv, enable Force Conform = true
  5. Remove Stream By Property (category: ffmpegCommand) -- keep only English audio:
    • Property To Check: tags.language
    • Values: eng,und
    • Condition: remove streams that do NOT match (keeps English + undefined)
    • Stream type: audio
  6. Remove Subtitles (category: ffmpegCommand) -- removes all embedded subs (Bazarr provides external subs)
  7. Remove Data Streams (category: ffmpegCommand) -- removes cover art, fonts, attachments
  8. Custom Arguments (category: ffmpegCommand) -- Output Arguments: -c copy (copy all streams, no re-encode)
  9. Execute (category: ffmpegCommand) -- runs the assembled ffmpeg command
  10. Replace Original File (category: file) -- moves the processed file from cache back to the source location

Wire Check File Medium's first output (video) to Begin Command. This flow remuxes files without re-encoding video or audio. It only strips unwanted streams.

3. Add Libraries

Go to the Libraries tab and click Library+ for each:

Movies: - Source: /data/media/movies - Transcode cache: /temp - Containers: mkv,mp4,avi - Transcode options: select the remux flow created above - Source options: enable Scan on start, enable Folder watch, set Process Library to ON - Schedule: 24/7 (or limit to off-peak hours)

TV Shows: - Source: /data/media/tv - Same settings as Movies

4. Start Processing

  1. Click Scan (Fresh) from the Library Options button to do a full scan.
  2. Monitor the dashboard -- files will move through the queue.
  3. Target status for completed files: "Transcode: Not required".

Dependencies

Dependency Purpose
Shared media (arr-data) Reads source files and writes transcoded output
Intel GPU node Required for hardware-accelerated transcoding

Upstream