Plex Media Server 1.42 – Buffer growth not trimmed during audio transcode

Plex Media Server
1

Server Version#:1.42
Player Version#: Current Apple TV
Summary
In Plex Media Server 1.42, when the transcoder temporary directory is mapped to a RAM disk (/dev/shm), memory usage grows continuously throughout playback when the video stream is direct-played and only the audio is transcoded. The transcode buffer never trims during playback, and the memory is not released after playback ends until the server or container restarts.

This behavior was confirmed with TrueHD audio being transcoded to AAC. It is unknown whether the same issue affects other audio codecs.

Environment

  • Plex Media Server 1.42.x (Linux / Docker build)

  • Transcoder temporary directory: /transcode → /dev/shm

  • Hardware transcoding: Intel Quick Sync (enabled)

  • Clients tested: Plex Web and Apple TV

  • Source: 4K H.265 video (direct play) with TrueHD audio (transcoded to AAC)

  • System memory: ~50 GB total, /dev/shm tmpfs ~26 GB

  • Buffer set to 60 sec in Transcode settings

Steps to Reproduce

  1. Map /transcode to /dev/shm and set “Transcoder temporary directory” to /transcode.

  2. Play a 4 K H.265 title with TrueHD audio that requires audio-only transcoding.

  3. Monitor memory usage during playback using df -h /dev/shm or free -h.

  4. Observe steady memory growth during playback.

  5. Stop playback and confirm that the memory is not reclaimed.

Expected Behavior
Plex should maintain a bounded buffer size consistent with the “Transcoder default throttle buffer” setting and release memory once playback ends.

Actual Behavior
Memory allocated for transcoding grows continuously throughout playback and is not freed until Plex is restarted. The throttle buffer setting appears to have no effect.

Additional Notes

  • Confirmed with TrueHD audio; untested with other audio codecs.

  • Behavior began after upgrade to PMS 1.42 — earlier 1.41 builds not confirmed.

  • Long sessions or multiple concurrent streams can exhaust tmpfs and cause playback errors or OOM conditions.

  • Behavior occurs during audio & video transcodes as well

  • Disk usage accumulates at about 20GB per hour of watch time when only transcoding audio and direct stream of video.

2

Can you enable Debug logging, repeat the steps, download your PMS logs, and share them with me?

Are you using the official Plex PMS Docker? Does this also happen when the transcode directory is mapped to the default location or another location? Are you willing to try to reproduce with our latest Beta 1.43.0?

Is this only happening when transcoding audio (TrueHD) or does this also happen with video transcodes?

3

Running plexinc/pms-docker from https://hub.docker.com/r/plexinc/pms-docker/
Can run an alternate repo is you recommend.

I enabled verbose logs, and ran a couple scenarios.

  1. played a DTS track, no RAM utilization, no transcoding.
  2. Played a TrueHD track with no video transcode, RAM utilization starts to collect right away during audio transcoding.
  3. Force video transcode during DTS playback, RAM utilization seems to build without leveling, but not as quickly as the audio only scenario.

If I upload the logs are they only available to Plex employees or can the community download?

1 Like
4

Debug logging, not verbose please. You can DM me the logs if you like.

5

Just tried to DM them to you.

An error occurred: Sorry, Atomatth is not accepting messages at the moment.

I can try the alternate build if you want to specify a repo.

Thanks for your help.

6

I messaged you so you should be able to share your logs with me. Are you still observing an issue?

7

I am facing this same issue with a Proxmox LXC container. I am running 1.43.0.10346. This was happening on 1.42.2.10156 as well.

8

Issue exists with version 1.43.0.10389 as well

9

This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.

10