High CPU usage + control packet loss (v3.12.1 Linux)

Hey guys,

we experience some strange behaviour at our 1024 slot teamspeak server.
The situation:

  • high cpu usage up to ~170% of a single core (on a Ryzen 9 3900X)
  • control packet loss up to aprox. 4-6%
  • delay in sending text messages, moving and mute/unmute up to 6 minutes
  • there is no significant traffic at the time of occurance (tx 3-5 mbit/s | rx 1-3 mbit/s)
  • there is no other process or service running on the server/cpu

What we did so far:

  • turned off encryption for ingame and public channels
  • we tried switching the server to a different hardware and different network in a hope of difference
  • also there is no attack or some external influence interfering, we monitor that as well of course

The clients of our server uses the feature of muting/unmuting themselfs as they come in range of each other to communicate, as a part of the game.

At the server log there even are dropped control packets:

2020-09-20 18:36:55.445380|INFO    |PktHandler    |33 |Cleaning up connection because of 15 resends of COMMAND packet
2020-09-20 18:36:55.445444|INFO    |PktHandler    |33 |Dropping client 26911 because of resend timeout
2020-09-20 18:37:07.711250|INFO    |PktHandler    |33 |Cleaning up connection because of 14 resends of COMMAND packet
2020-09-20 18:37:07.711294|INFO    |PktHandler    |33 |Dropping client 26077 because of resend timeout
2020-09-20 18:37:10.230523|INFO    |PktHandler    |33 |Cleaning up connection because of 14 resends of COMMAND packet
2020-09-20 18:37:10.230554|INFO    |PktHandler    |33 |Dropping client 26481 because of resend timeout
2020-09-20 18:37:11.744644|INFO    |PktHandler    |33 |Cleaning up connection because of 14 resends of COMMAND packet
2020-09-20 18:37:11.744687|INFO    |PktHandler    |33 |Dropping client 25857 because of resend timeout
2020-09-20 18:37:17.196284|INFO    |PktHandler    |33 |Cleaning up connection because of 14 resends of COMMAND packet
2020-09-20 18:37:17.196374|INFO    |PktHandler    |33 |Dropping client 26723 because of resend timeout
2020-09-20 18:37:23.415269|INFO    |PktHandler    |33 |Cleaning up connection because of 14 resends of COMMAND packet
2020-09-20 18:37:23.415341|INFO    |PktHandler    |33 |Dropping client 26760 because of resend timeout

We do not recognize patterns or situations related to the peaks of overload that last between 2min and 30min like explained above.
Unfortunately, it’s not comprehensible to us what causes or resolves it again.

Does anyone have any ideas, informations, experience with such situations?
I’m willing to provide any information that could help investigating the problem.

We would be happy for any comments :slight_smile:

3 Likes

3 posts were merged into an existing topic: Querypoolsize paramter is not recognized / Fivem server is laggy

A post was merged into an existing topic: Querypoolsize paramter is not recognized / Fivem server is laggy