SwarmAgent [UpdateTaskState]: Task Killed remotely

Development Programming & Scripting Multiplayer & Networking
, , , , ,
1

I’m trying to bake lightmaps for my Unreal Tournament level over my network with 2 other pcs.

The client pcs are CONSTANTLY restarting the process at completely random intervals making the remote clients barely complete any work.

Before each interuption, the log of the host pc spits out:

9:02:59 PM: [UpdateTaskState]: Task Killed remotely by DS-TV
9:02:59 PM: [UpdateTaskState]: Requeueing: SMTextureMapping
9:02:59 PM: [UpdateTaskState]: Task Killed remotely by DS-TV
9:02:59 PM: [UpdateTaskState]: Requeueing: SMTextureMapping
9:02:59 PM: [UpdateTaskState]: Task Killed remotely by DS-TV
9:02:59 PM: [UpdateTaskState]: Requeueing: SMTextureMapping
9:02:59 PM: [UpdateTaskState]: Task Killed remotely by DS-TV
9:02:59 PM: [UpdateTaskState]: Requeueing: SMTextureMapping
9:02:59 PM: [UpdateTaskState]: Task Killed remotely by DS-TV
9:02:59 PM: [UpdateTaskState]: Requeueing: SMTextureMapping
9:02:59 PM: [UpdateTaskState]: Task Killed remotely by DS-TV
9:02:59 PM: [UpdateTaskState]: Requeueing: SMTextureMapping
9:02:59 PM: [UpdateTaskState]: Task Killed remotely by DS-TV
9:02:59 PM: [UpdateTaskState]: Requeueing: SMTextureMapping
9:02:59 PM: [UpdateTaskState]: Task Killed remotely by DS-TV
9:02:59 PM: [UpdateTaskState]: Requeueing: SMTextureMapping

Followed shortly after by the client restarting the process.

What is going on?

2

Not enough memory maybe?

I don’t remember the specific error but a year or so ago I had a similar problem with a remote lightmass constantly restarting jobs. Adding more memory to that machine fixed the problem.

When I had that problem, the machine had 4G of RAM (1G per core). It only happened on very large maps and demos containing a lot of complex materials and meshes.

I believe they reduced lightmass’ memory useage since then.

IMO you want at least 1G of RAM per core to run lightmass, but 2G is better. Not sure how hyper-threading affects that, might need more memory with it.

3

The remote systems have 8GB, and the host has 24GB. The level it is lighting does not have particularly complex geometry or materials either.

Probably not a memory issue, but if something is happening on the clients they wipe the log each time they restart the process making it impossible to tell vOv