UE 5.6.1 causing extremely high UDP traffic

Since around the 25th of September this year we have several classroom machines running UE 5.6.1 which constantly spam UDP messages through the network. Around 97% of our total network capacity is being consumed by this traffic, around 9 gbit, which is of course causing quite a lot of problems.

When tracing the packages it seems the machines are sending ChaosVDSessionPing and ChaosVDSessionPong messages between one another at the rate of about a 1000 a second per machine.

This does not happen on all machines running this particular version but the majority of the 50 or so users that run UE daily have this problem. When they all shut down UE the traffic stops like clockwork.

At the same time this error appeared nearly every time they save their work:

/Game/OurLevels/Level_1 Validating asset
/Game/OurLevels/Level_1 Validating asset
/Game/OurLevels/Level_1 Validating asset
/Game/OurLevels/Level_1 Warnings logged while validating asset FUdpMessageProcessor::UpdateReassemblers Discarding 2/4 of stale message segements from Sender=10.xxx.xxx.xxx:57559, Sequence=0
FUdpMessageProcessor::UpdateReassemblers Discarding 3/4 of stale message segements from Sender=10.xxx.xxx.xxx:57559, Sequence=0
FUdpMessageProcessor::UpdateReassemblers Discarding 3/4 of stale message segements from Sender=10.xxx.xxx.xxx:57559, Sequence=0
FUdpMessageProcessor::UpdateReassemblers Discarding 3/4 of stale message segements from Sender=10.xxx.xxx.xxx:57559, Sequence=0
FUdpMessageProcessor::UpdateReassemblers Discarding 3/4 of stale message segements from Sender=10.xxx.xxx.xxx:57559, Sequence=0
FUdpMessageProcessor::UpdateReassemblers Discarding 3/4 of stale message segements from Sender=10.xxx.xxx.xxx:57559, Sequence=0
FUdpMessageProcessor::UpdateReassemblers Discarding 3/4 of stale message segements from Sender=10.xxx.xxx.xxx:57559, Sequence=0
/Game/OurLevels/Level_1 Validating asset
/Game/OurLevels/Level_1 Validating asset
/Game/OurLevels/Level_1 Validating asset
/Game/OurLevels/Level_1 Validating asset

which we are not sure is related but it could be, and started happening at the same time.

Any help or suggestion would be appreciated. We do not run any in-house plugins or other code, nor do we build the engine ourselves. This is stock standard UE from the EGL.

Hi,

I think this is related to an issue where the Chaos Visual Debugger remote session discovery system was active from editor start up instead of only being active when CVD (for short) is open.

CVD sends the session discovery ping once per second, and target applications (other editors, game servers and game clients) only answer when they get the ping. But I do see how the traffic can increase exponentially with multiple editors open in the same LAN, as they will be discovering each other.

A fix for that problem was made a few months ago, and will be available in UE 5.7.

The system doing the ping is not part of the ChaosVD plugin, so disabling it will not fix the problem either sadly.

https://github.com/EpicGames/UnrealEngine/commit/2e0db326b11c298ca8a3d57b3d82293ae88e8b05

As for a workaround, the only one at the moment is disabling the UDP Messaging plugin, but this is not an issue with that plugin. Disabling that plugin might/will break other functionality that depends on it. For example, in the CVD side, CVD will no longer be capable of discovering a packaged game client or server running in the same machine or LAN (which for this case seems like an acceptable trade off). Other tools might be affected differently.

Apologies for the issue