Jump to content

Emby 4.10.2.0 leaves orphaned Live TV stream after overlapping Samsung playback requests


Recommended Posts

hiimcourty
Posted

Emby Server 4.10.2.0 on Windows 11, Samsung Emby 2.3.6, M3U Live TV tuner with Rebel IPTV.

 

October 25, 2026 (Pacific): At ~6:43:28 AM overlapping Samsung TV playback-info requests for Food Network created TWO consumers of the same SharedHttpPipelineSource. Emby log reported consumer count 2. At 7:50:08 AM the tracked playback went idle, was stopped and removed; consumer count only fell to 1. The second consumer was not subsequently removed. Hours later Emby Sessions API showed zero NowPlaying sessions, but EmbyServer.exe still had its original established TCP connection to the M3U backend from 6:43 AM. Rebel showed Food Network active even as its total viewer metric fluctuated between 0 and 1. Stopping Food Network in Rebel did not permanently stop the stream. Restarting Emby at ~11:43 AM cleared the orphan, with Rebel showing 0 viewers and 0 feeds.

 

Expected: concurrent playback-info calls should not create a permanently untracked consumer; all upstream streams should close when their legitimate playback sessions end. Is this a known SharedHttpPipelineSource consumer-count bug? Is there a supported workaround without modifying Emby's build or disrupting current viewers?

 

Can supply anonymized logs privately upon request.

hiimcourty
Posted

Thanks, Luke. I have the complete archived Emby Server log from October 25, 2026 (pre-restart), but it includes private hostnames, IP addresses, URL/query parameters, internal identifiers and local file paths. Rather than uploading that raw log publicly, here is a targeted, anonymized extraction of the exact playback/consumer lifecycle events. All timestamps are Pacific Daylight Time (UTC-7), October 25, 2026.

 

06:43:22.719 LiveTV: Streaming Channel [Food Network], with streamId: STREAM_1

06:43:28.792 SharedHttpPipelineSource: LiveStream consumerId added: CONSUMER_A

06:43:28.792 SharedHttpPipelineSource: Live stream STREAM_1 consumer count is now 1

06:43:28.792 SharedHttpPipelineSource: LiveStream consumerId added: CONSUMER_B

06:43:28.792 SharedHttpPipelineSource: Live stream STREAM_1 consumer count is now 2

06:43:28.792 SessionManager: Setting _lastPlaySessionId for session [REDACTED] to CONSUMER_A

06:43:28.793 SessionManager: Setting _lastPlaySessionId for that SAME session to CONSUMER_B

06:43:30.292 SessionManager: Playback start reported by app Emby for Samsung 2.3.6 on Samsung Smart TV playing Food Network. PlaySessionId: CONSUMER_B

07:50:08.333 SessionManager: PlaySession CONSUMER_B has gone idle while playing

07:50:08.333 SessionManager: Playback stopped reported by app Emby for Samsung 2.3.6 on Samsung Smart TV playing Food Network. PlaySessionId: CONSUMER_B

07:50:08.333 SharedHttpPipelineSource: LiveStream consumerId removed: CONSUMER_B

07:50:08.333 SharedHttpPipelineSource: Live stream STREAM_1 consumer count is now 1

11:42:47.917 MediaSourceManager: Closing live stream [REDACTED]_STREAM_1 (on Emby shutdown)

11:42:48.290 MediaSourceManager: Live stream [REDACTED]_STREAM_1 closed successfully

 

A search of the complete pre-restart server log finds CONSUMER_A being added and set as _lastPlaySessionId but no matching consumer removal before shutdown. Emby Sessions API reported zero NowPlaying users while EmbyServer.exe held its original HTTP connection to the IPTV backend, established at 06:43:22. The IPTV backend continued reporting Food Network active; manually stopping it there did not keep it stopped. Restarting Emby at ~11:43 AM released the source immediately; the backend then showed zero active feeds/viewers. No reproduction or configuration changes have been attempted since.

 

I also checked the available ffmpeg/transcode/remux/directstream log directory and found no new ffmpeg logs dated October 25 for this incident (the newest of those logs were from October 24).

 

What secure/private method would you prefer for providing the FULL complete server log, including only a dashboard-downloaded sanitized version if that is sufficient? I don't want to expose private URLs, hostnames, access parameters, household names, or local paths publicly. The original archived pre-restart log is retained.

  • Thanks 1
Posted

OK we are looking into it. Thanks.

  • 3 weeks later...
hiimcourty
Posted

Hi Luke, Just following up on this issue. The same behavior happened again today, November 11, this time on a Google TV as well. Do you have any updates on the investigation or a potential fix? Thanks!

Posted

HI, can you please PM me the complete server log? Thanks.

awkimball
Posted

Seeing this exact same issue on Emby Server 4.10.2.0 (Debian Linux x64) with an M3U tuner, and I have full sanitized server and ffmpeg logs from a clean repro attached in case that helps track it down.

In our case a Roku client (Roku SG 4.1.70) tuned MS NOW at 13:37:05 and retried twice after hitting a DirectPlayError on Direct Stream before falling back to transcoding:

  • 13:37:05 - Attempt 1 (PlaySessionId a80dc1343fe84496816702074cd9c579) opens live stream b0b0efceecc72a60e133c4cea9444d7d (consumer count 1) and sends Playback start. When it goes idle 60s later at 13:38:09, its consumer is removed normally.
  • 13:37:09 - Attempt 2 (PlaySessionId cfa999bb1306430093cacb37781535de) calls PlaybackInfo (consumer count 2), hits DirectPlayError on the first HLS segment before ever sending Playback start, and immediately retries as Attempt 3 at 13:37:12 (PlaySessionId d9edc7a1361d49da9b436798c80c1ac1, consumer count 3).
  • 13:38:12 - When Attempt 2 times out after 60s, SessionManager logs "PlaySession cfa999bb1306430093cacb37781535de has gone idle while playing", followed by Playback stopped with an empty PlaySessionId ("PlaySessionId: ."), and never removes cfa999bb1306430093cacb37781535de from SharedHttpPipelineSource.
  • 17:04:49 - The user stops watching and exits, which drops consumer count from 2 to 1 instead of 0. SharedHttpPipelineSource kept pulling the upstream M3U stream with zero active viewers for nearly 5 hours until I manually called POST /LiveStreams/Close at 21:58:50.

We also saw the same leak earlier with an iOS client that requested PlaybackInfo, fetched two segments, and disconnected before sending Playback start.

Sanitized embyserver.txt (with full startup header and unedited 13:34-13:39, 16:58-17:05, and 21:58 windows; steady-state HLS segment polling in between trimmed for upload size) and the three ffmpeg logs from 13:37 are attached. Thanks!

@Luke

ffmpeg-transcode-05ef2409-b039-4307-b69e-e60d5e08b38e_1.txt ffmpeg-directstream-e23d5fdb-4313-4479-a1f9-82c7641e4514_1.txt ffmpeg-directstream-e2b12d83-14d9-46b9-9aaa-c16f46f43096_1.txt embyserver.txt

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
×
×
  • Create New...