[7.1.16 - 1.7.10] Builder causes players to time out
gioporta opened this issue ยท 21 comments
A player on our server put a BC builder together and now anyone who gets near the area consistently times out, and the serverside logs show nothing of importance except that the player "disconnected"
Might this have anything to do with running the server on a Haswell processor? I'm going to try moving the server to an Ivy Bridge machine.
For the record, it's running on a Google Cloud VM. Issues don't seem to be present in single player.
no the processor shouldn't have anything to do with it on it's own, most likely the server just can't keep up with everything that is going on
@AEnterprise if you count "everything that is going on" as 1 person using a builder/quarry/whatever, then I find that hard to believe, especially since it works fine in singleplayer on a computer with inadequate specifications.
@AEnterprise you asked this already.
yes but i rered that and you say it's a VM with 8GB, that doesn't mean you allocated 8GB to minecraft, i'm basicly asking to verify how much is allocated
Do the clients have any information (crashes, errors or other problems) in their logs?
No, just a network timeout in the client log.
[09:41:15] [Netty Client IO #4/ERROR] [FML/]: NetworkDispatcher exception
io.netty.handler.timeout.ReadTimeoutException
Basically what happens is the client goes near the Buildcraft builder. Everything is working fine, and you can see the builder "spitting out" blocks as it is building. But a few seconds later everything stops, and then you time out. I also recorded CPU spikes of around +80% for the server when this was happening.
The builder was tested with the same blueprint in singleplayer and there were no issues.
can you give a full client log of this heapening and a server one to match? might be more info about the cause in there
The logs are way too big to paste, so I'll just upload snippets.
Relevant part of the server log - shows time outs of two clients
Client log from join to time out - time out @ 09:41:15
Abnormal CPU usage on the server during the entire ordeal:
so it looks like the cpu usage spikes to 100% when this heapens that means the timeouts are not directly buildcrafts fault but your server getting overloaded the client to not get it's response packets in time causing the timeout
what are the server specs? it might not have enough ram allocated
I highly doubt it's an issue with the server specs.
Server is a VM with 8GB of RAM and 2 vCores of a Xeon E5-2670v3.
I can run robots and build with those without spiking the CPU whatsoever, but the moment a builder block comes into play, it causes issues.
Yeah its possible that its a large blueprint and we try to do too much checking in a single tick. Or perhaps its too much traffic being generated.
The same issue occurs with the quarry so it has nothing to do with blueprints. The moment you turn it on the TPS slows down to a halt and the CPU usage is all over the place.
is this also a problem on 1.8? if not this can be closed
maybe also resolved with the 8.0 rewrite?
I've noticed this problem as well, but with the quarry as Giovanni mentioned. Since it's a common issue for those two blocks, could it somehow be related to the red deletion blocks that they spit out?
Both the quarry and the builder use the same blueprint builder code (the quarry just generates a blueprint for the shape and size of the quarry, and it's the same with the filler for all its different types).
The red breaking blocks are an animation, and are done in the same way that other blocks going out of the blueprint are. I'm not sure how they could be todo with it though, other than showing that the issue occurs when it's building its blueprint.
One possibility that came up is that VeinMiner, more specifically version 0.34.1, is the one causing this (it didn't happen on 0.31.6 when I went back to it, at least for the quarry). Do you have that mod in your pack?