running a local headless in an Ubuntu 18.04 VM (4 core x 12GB) running on a dell poweredge 710. It is on a different vlan than my game client (specs in log file) so I have to connect by ip:port. This is hosted for online multiplayer with some folks newer to the game and shows up in the multiplayer menu. This server has 0 mods and is a direct install from factorio's download page
The blah blah:
I've been using the same computer for years, I don't think I've played Factorio as much on any other PC. I only changed the GPU from a 970 GTX to a 1080 TI a few years ago (before the whole bitcoin bubble). I've also hosted another multiplayer in the past year on this same server, I don't specifically recall if that factory was larger or not. I did some searching around on the forums using the keyword stutter and found many hits. After going through a few pages of folks on the same version I did not find anything that felt applied to my situation so I'm making my own post.
Details:
I've been having stuttering. Most notably when using Magic Shoes (what I call the exoskeletons). With the debug menu open I see the FPS/UPS going from 60 to 59 and the wait icon (stopwatch) that is next to it flicker when doing so for every stutter. I've attached logs, screens, and if there is a good place/suggestion to host the map if necessary I can do so. Otherwise I could PM the password of the multiplayer map to a Factorio employee
.I have ran into laggy issues before. Years ago, a friend and I found that it seemed to be with using too much stone/concrete on the ground. Maybe that was never the case, but when played without it things were fine.
This issue "feels" like an issue I recall having previously, a super-bad train path somewhere. Bringing either the client or server to it's needs. However in my testing it "seems" to be a lot better (but still occurs) when I set all of the trains to manual and "occurs more often" when the trains are running (there are only like 4-5 trains total, 1 route each, nothing on the same route between stations), making the "take a screenshot when good" harder to try to time.
Also, there seems to be very large (my own relative assumptions) of pathfinder "steps" and "work" done per tick. You can see them in the screenshots but they can burst from <= 1 to 7/8000. That might not be relevant at all and just a coincidence. While the wait icon flashes "seem" to occur somewhat close to those spikes, it is not at the exact same time.
Thanks for your help. I tried to follow the performance reporting guidelines best I could. Let me know if I missed something.
-Red
