[0.17.79] Map download fails if maps larger than 2MB

[geek-at]

I've had an interesting problem the last couple of days and couldn't find any solution on the forums or borad google searches.

I'm (network engineer as day job) hosting a dedicated server in my LAN to play with my wife. We have played on LAN over 700 hours and large maps without problems.

Recently a friend of ours wanted to join so I set up port forwarding and we created a new map. He was able to join without a problem and we played 3 hours that day.
Next day he tried to join and the download didn't finish for him. It started but then stopped at quarter or half download. On the third try it worked so we ignored it.

By now the map had 10 hours of play time and now he has the downloading problem every time he tries to connect. We can still connect from LAN without a problem.



Here's what I know:

• The larger the map download is, the less likely any external user can finish the download (and therefore connect)
• The first second or so the download works but then the B/s and ETA drop to 0 and nothing happens
• 3 more friends have tested it and have the same problem
• I can't reproduce the error from within my network but when using a VPN and connecting via that, I have the same problem
• I'm using a PFsense box so I set up outgoing static port mapping
• The download never stops at the same download size. Sometimes it stops with 200kb, sometimes it gets to 4mb (almost finished)
• If the download completes (happens increasingly rareley (depending on the map size), anyone can play without a problem
• If I delete the map and restart the server, everyone can connect (because download is so small it finishes for everyone)
• I'm using a business fiber internet connection and have multiple static external IPs so no blocking or rate limits apply
• Nothing changes if I change the "segment_size" settings in the server config to higher or lower values
• The Factorio server is running from a Docker container with --net=host

Relevant log snippet (doesn't output any errors)

151.684 ConnectionRequestReplyConfirm (type(ConnectionRequestReplyConfirm) ) from(IP ADDR:({82.218.xx.117:59546}))
151.684 Info ServerRouter.cpp:425: Replying to connectionRequest for address(IP ADDR:({82.218.xx.117:59546})).
151.684 Info ServerSynchronizer.cpp:599: nextHeartbeatSequenceNumber(4337) adding peer(40)
151.703 Info ServerMultiplayerManager.cpp:776: updateTick(1763745) changing state from(InGame) to(InGameSavingMap)
151.737 Info ServerMultiplayerManager.cpp:925: updateTick(1763745) received stateChanged peerID(40) oldState(Ready) newState(ConnectedWaitingForMap)
152.324 Info ServerMultiplayerManager.cpp:993: UpdateTick(1763745) Serving map(/opt/factorio/temp/mp-save-38.zip) for peer(40) size(5150412) crc(2617383273)
152.324 Info ServerMultiplayerManager.cpp:776: updateTick(1763745) changing state from(InGameSavingMap) to(InGame)
152.375 Info ServerMultiplayerManager.cpp:925: updateTick(1763745) received stateChanged peerID(40) oldState(ConnectedWaitingForMap) newState(ConnectedDownloadingMap)
153.302 Info ServerMultiplayerManager.cpp Disconnect notification for peer (40)
153.302 Info ServerMultiplayerManager.cpp:925: updateTick(1763745) received stateChanged peerID(40) oldState(ConnectedDownloadingMap) newState(DisconnectScheduled)
2020-05-03 21:17:37 [LEAVE] r3d4rr0w left the game

Any pointers on what else I could try?

[Loewchen]

I'd assume a firewall issue. The client log would tell you more.

[Zavian]

viewtopic.php?p=492688#p492688 sounds potentially similar. (If I recall correctly, other people have also reported that they have resolved multiplayer connection issues by limiting bandwidth in the server config file).

If you have time and inclination it might be worth getting packet traces of connecting via vpn to try to work out the root cause of the issue. (That disconnection message 1 second after the client starts downloading the map just seems weird, unless the network stack is reporting some sort of error that causes Factorio to think the client is gone/unreachable, or that otherwise causes Factorio to think it can't recover that client connection, so it just disconnects that client).

Other potentially similar cases that come to mind.

Some people have had issues with PFSense NAT rules in the past. eg viewtopic.php?t=47110
(But I doubt that is your problem, since then it would probably have never worked for the first session).

A year or so ago someone reported an issue which was eventually tracked to the map download setting off some sort of network monitoring/intrusion detection system on their corporate network that blocked/killed the connection.

Edit: Might also be worth checking whether you can reproduce it with the latest experimental version. (0.18.22)