Server Version#:1.18.1.1973
Player Version#:Latest Roku, Latest PMP, Latest Web Player
On versions 1.18.0.1846 through 1.18.1.1973 I have found that seemly random videos will end early 2-5 minutes. It appears to be when the transcoding finishes. I am using PMS on ubuntu server and hw transcoding on a p400. In the zip file i attached, this happens to user McCalebs on Legend of Korra episode Long Live the Queen at Oct 17, 2019 13:48:46.139 in Plex Media Server.log
Plex Media Server Logs_2019-10-17_13-49-39.zip (9.5 MB)
Can anyone acknowledge this?
Sorry for missing this post.
Looking at your logs shows:
- Transcoder running along very nicely and going into sloth mode (it’s ahead of what’s needed so sleeping). 398 transcode buffers were processed properly.
- I see the request as if on a public IP via a VPN
172.20.0.9 via a 207.x.x.x address.
The disturbing part here is 172.20.x.x is part of the RFC-1918 compliant addressing block for Private Network and should not be an issue.
Here you can see that the processing was ok then the player stopped without notification but started another.
Is this confirming what you experience?
Oct 17, 2019 13:20:08.991 [0x7fb52dffb700] DEBUG - Transcoder segment range: 0 - 400 (399)
Oct 17, 2019 13:20:09.237 [0x7fb5d7fff700] DEBUG - Completed: [172.20.0.9:44398] 200 GET /video/:/transcode/universal/session/67a40a1d-031a-49c7-bd86-de69b4efab4a-4/base/00001.ts?protocol=hls&mediaIndex=0&waitForSegments=1 (11 live) TLS 445ms 1148304 bytes (pipelined: 1)
Oct 17, 2019 13:20:09.237 [0x7fb5d7fff700] DEBUG - Removed transcode data consumer, active count 1 => 0
Oct 17, 2019 13:20:09.578 [0x7fb5d7fff700] DEBUG - Auth: authenticated user 1 as LrrrAc
Oct 17, 2019 13:20:09.578 [0x7fb52cff9700] DEBUG - Request: [172.20.0.14:52268 (Subnet)] GET /status/sessions (12 live) GZIP Signed-in Token (LrrrAc)
Oct 17, 2019 13:20:09.578 [0x7fb52cff9700] DEBUG - [Now] Adding 2 sessions.
Oct 17, 2019 13:20:09.579 [0x7fb5d7fff700] DEBUG - Completed: [172.20.0.14:52268] 200 GET /status/sessions (12 live) GZIP 1ms 2380 bytes
Oct 17, 2019 13:20:09.627 [0x7fb5dcef4700] DEBUG - Using X-Forwarded-For: 207.242.237.50 as remote address
Oct 17, 2019 13:20:09.628 [0x7fb5dcef4700] DEBUG - Auth: authenticated user 1 as LrrrAc
Oct 17, 2019 13:20:09.628 [0x7fb4edffb700] DEBUG - Request: [172.20.0.9:44424 (WAN)] GET /:/timeline?ratingKey=55088&key=%2Flibrary%2Fmetadata%2F55088&playbackTime=2227841&playQueueItemID=287097&state=paused&hasMDE=1&time=2228000&duration=2703000 (13 live) TLS Signed-in Token (LrrrAc)
Oct 17, 2019 13:20:09.630 [0x7fb4edffb700] DEBUG - Client [optm5v9xjdrkerpp95mj5fx8] reporting timeline state paused, progress of 2228000/2703000ms for guid=, ratingKey=55088 url=, key=/library/metadata/55088, containerKey=, metadataId=55088, source=
Oct 17, 2019 13:20:09.631 [0x7fb4edffb700] DEBUG - Play progress on 55088 'Sawbones: Rock Medicine' - got played 2228000 ms by account 1!
Thanks for the response @ChuckPa Chuck! Yeah it more or less does that. It’ll end early and start the 10 second countdown for the next show right away. It seems to happen only when using hardware transcoding with the new decoding updates.
@ChuckPa Just checking in to see if theres any update. I was happy when hardware decoding was added to plex after working on the workaround patch for a bit but this is making me want to go back to the patch. Is there anything I can do?
I have my plex on a windows machine, but i’m also experiencing this. Every time a stream plays, it ends a few mins early and auto skips to the new one. Very frustrating, the thing is, it’s not happening on all devices on my network, seems only the ones using WiFI – the hardwired Rokus i have don’t experience this.
That makes it a Player problem or a WiFI problem?
If truly a player problem, I can’t be of any help. I am only now starting to learn more about the players.
WiFi is tricky. If 2.4 Ghz band, anything could make it stop at any point. It might be running out of bandwidth / the handshaking might get interrupted. Debugging WiFi, without being on site to see and test signal strength & bandwidth makes it difficult.
What happens if you install a speedtest app on the WiFi device(s). Do you get full, sustained wifi speed?
Ive had this happen both on a roku with 48Mbps down ten minutes away from the server that has 1Gbps upload, and a laptop across the state that gets throttled down to 1.5Mbps. Both wifi, but it doesnt seem like the exactly 5 minutes before the end of the show should happen due to network issues.
No, I agree. If it’s exactly 5 minutes before the end of the playback, that has to be a player problem.
I am attempting now with Shield 2019, Roku 3 and iDevice.
I’ve added the shield since my last attempt
I am also using QSV and not an nVidia card.
Can you spin me a new set of log files?
Also, which nVidia driver version do you have installed?
Thanks for testing! Ive tried with several, 375.39, 430.14, 440.31, 440.36. Ill try to get logs.
Also the issue goes away completely when i disable HW transcoding.
The minimum nVidia version we need is 418.30
We need to find out what is uniquely going on with yours.
If it were widespread, we’d hear all about it. The nVidia cards are really popular.
im currently running 430.14
Good. thanks for that.
Would you mind making sure DEBUG is on, VERBOSE is off.
and recreating?
As soon as it’s over, grab the ZIP file and attach.
Its not consistent, but i started after you asked and just got one to skip 4 minutes into the show. Heres the logs: Plex Media Server Logs_2019-12-05_21-47-46.zip (6.2 MB) The show was Brooklyn Nine Nine: Yippie Kayak
Your logs aren’t showing anything other than it transcoding, idling, transcoding.
This really looks like the player is misbehaving but I can’t be definitive.
I certainly don’t see a server problem.
I do keep seeing an external address (with X-Forwarder) info.
Is this a VPN ?
Im not using a vpn on anything. The server is external to the network however. Also ive had reports of this happening on both roku and the web app. And only when I have hardware transcoding enabled. Im not sure where to go from here.
Can you definitively isolate HW transcoding versus SW transcoding?
I ask because with the Ryzen, you’re using the external GPU card.
while it should be zero-copy transcoding (transcodes directly to the output buffer),
if blocks are missing then that’s a problem.
Intel QSV and Software transcoding operate differently.
QSV is zero copy and software is a 1:1 read from input -write to output buffer.
The difference here signalling from the nVidia libraries that the buffer is ready versus software signalling.
Yes, I have never had any issues with SW transcoding and my Ryzen 2600 has no option for QSV obviously. I have had users say it happens quite often and turn off HW transcoding and they say the issue goes away. Ive also tried several times to turn it back on as it takes away a significant amount of power draw, and had to turn it back off due to the annoyance of having to back up an episode and scrub through to find where I was.
Are there logs that I can try to gather on the nvidia end of things?
