Weird Transcoding Issues

Server Version#: 1.14.0.5470
Player Version#: Roku 5.3 Build 10
Vmware ESXi 6.5 - Plex VM Running on Ubuntu Linux 16.04LTS fully updated
8 vCPU’s - 4Gb Memory (Intel Xeon e5-2960)

Im having a very strange problem with VC-1 files on my Plex Server. It seems like this just started recently and I haven’t been able to figure out what the issue is. When I play VC-1 files my server will transcode the media, this is a default behavior, Ive never had an issue with this working until recently. Now my stream will constantly buffer and looking at the logs it shows…

Dec 05, 2018 14:32:52.812 [0x7fed217ff700] VERBOSE - * speed => 0.3

But other times the streaming works just fine with speed number between 1.9-3.3. When this was working without issues (earlier today) I started 3 x VC-1 transcodes simultaneously and all had no issues. But then randomly I cant even play a single VC-1 file without buffering every 30 seconds.

Here are the things I have tried.
-Built new plex server in Ubuntu 16.04LTS (was on 18.04LTS)
- First time using the new server transcode speed numbers were fine with no buffering. Mid stream
the speed numbers decreased and buffering started. (20 min or so into the movie)

  • Changed the throttle buffer to 180
  • Tested network speed 400+ Mbps to each station having issues
  • Turned off all other VM’s on ESXi host
  • Performance tested the CPU’s in my machine, single threaded and multi threaded and get really good results (see below)

jhadmin@PLEX:~$ sysbench --test=cpu --num-threads=1 run
sysbench 0.4.12: multi-threaded system evaluation benchmark

Running the test with following options:
Number of threads: 1

Doing CPU performance benchmark

Threads started!
Done.

Maximum prime number checked in CPU test: 10000

Test execution summary:
total time: 8.8823s
total number of events: 10000
total time taken by event execution: 8.8764
per-request statistics:
min: 0.86ms
avg: 0.89ms
max: 2.82ms
approx. 95 percentile: 0.93ms

Threads fairness:
events (avg/stddev): 10000.0000/0.00
execution time (avg/stddev): 8.8764/0.00

jhadmin@JH-SRV-PLEX:~$ sysbench --test=cpu --num-threads=8 run
sysbench 0.4.12: multi-threaded system evaluation benchmark

Running the test with following options:
Number of threads: 8

Doing CPU performance benchmark

Threads started!
Done.

Maximum prime number checked in CPU test: 10000

Test execution summary:
total time: 2.4756s
total number of events: 10000
total time taken by event execution: 19.6115
per-request statistics:
min: 0.96ms
avg: 1.96ms
max: 52.91ms
approx. 95 percentile: 12.97ms

Threads fairness:
events (avg/stddev): 1250.0000/68.20
execution time (avg/stddev): 2.4514/0.02

Im going mad trying to figure this out, I was going to try to downgrade Plex but I cant find any early distributions online. Im not sure where to go from here, what else can I look at? Ive attached the logs files from the stream when it was not working and also when it was working. Same machine, same everything. This last time I got transcoding to start working again by just rebooting the machine, but that doesnt always fix the issue…

Any help or other things to try would be appreciated.

Update

Its not just VC-1 Files, Im trying to transcode a 4K stream down to 20Mbps and my trancoder speed numbers show 0.0 and it will never play. This also isnt only on the Roku Player, web client does the same thing. Again randomly works/then doesnt.

plex-broken-logs.txt (103.7 KB)

plex-working-logs.txt (84.4 KB)

May I have the full log file please?

It’s impossible to see what’s happening from that excerpt.

Also, how many VCPUs have been allocated to the VM?

Currently I have 8vCPUs assigned to the Plex Guest VM. Ive been messing with how many sockets / cores per socket to see if that fixes anything. No improvement though, still weird issues with sometimes transcoding working perfectly and others I cant do any transcoding. At this point I still have every other VM on the Host turned off.

Ive also turned all my Vmware Power Management settings to High Performance but have not tested yet.

Plex Media Server Logs_2018-12-05_14-18-04.zip (5.1 MB)

Would you mind turning VERBOSE back off?
DEBUG logging is the only thing we can really use at this point. VERBOSE generates so much output that we see about 2 minutes of elapsed time.

  1. Turn VERBOSE logging off
  2. Turn DEBUG logging on
  3. Recreate the issue
  4. Stop Playback
  5. Wait 30 seconds
  6. Capture the ZIP again
  7. Attach the ZIP

Update: I might have solved my own issue by changing Vmware Power Management settings from Balanced to High Performance. Ive been testing after the change and I cant recreate the issue. Multiple transcodes are happening at once with no problems. Ive rebooted/shutdown the host and guest multiple times and started/stopped the video playback with no issues.

Very odd. I did upgrade ESXi 6.0 to 6.5 a couple months ago, maybe this caused the sudden issue that I just hadnt noticed until now. Either way, Ill update with the logs you requested if it happens again. Thanks for your effort.

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