Linux dedicated server stops handling connections after long periods of running

Development Platform & Builds
, , , ,
1

Hi,

I have ran into a very strange issue which is preventing players from connecting to a dedicated server after 5-6 days of running continuously. This happens on a shipping build, source built 4.19.2. A reboot of the process resolves the issue and players can connect again. I have checked CPU and RAM, both are absolutely normal with no evidence of locking or resource limitations. Also tried on a dedicated server, and VPS.

When I attempt to connect to the server when it is in this non-responsive state, it logs the following periodically until the client receives a “connection time-out” error.

[2018.08.05-07.58.57:497][129]LogNet: NotifyAcceptingConnection accepted from: HIDDENIP:62315
[2018.08.05-07.58.58:500][169]LogNet: NotifyAcceptingConnection accepted from: HIDDENIP:62315
[2018.08.05-07.58.59:503][209]LogNet: NotifyAcceptingConnection accepted from: HIDDENIP:62315
[2018.08.05-07.59.00:506][249]LogNet: NotifyAcceptingConnection accepted from: HIDDENIP:62315

So I assume that the server is getting stuck around this area. I checked the source where this is being logged (https://github.com/EpicGames/UnrealEngine/blob/f509bb2d6c62806882d9a10476f3654cf1ee0634/Engine/Plugins/Online/OnlineSubsystemUtils/Source/OnlineSubsystemUtils/Private/IpNetDriver.cpp#L396).

I am trying to figure out what the correlation between server running duration and the failure of connections.

Any help would be appreciated.

Cheers,
Joshua

2

After further investigation I believe it has something to do with the handshaking process. I am still yet to find the correlation between server runtime and failing handshakes. Any help would be appreciated.

3

Have you ever found a fix for this problem? Could you share the solution if so?

4

I see the same issue with 4.24.3 dedicated server. If there is a solution, please share…

5

Hi there - we ended up setting up automatic reboots using the ‘Restart Game’ node in the Game Mode blueprint. This basically reloads the map and restarts all the net drivers and such along with it.