Hello, i run into a really annoying issue - it’s really curious that nobody else has noticed it yet…
I have a small Server running with 5 slots on a not so good PC but it works good. I normally update SpongeForge every 1-2 weeks. And have very few mods running like nucleas, permissionmanager and the dynmap.
I noticed that since a few days sometimes the CPU load runs at 50%, i thought it’s just the dynmap has a rendering issue but as i looked into the problem i tried to run a completly clean server with new map, new config, no mods and as i joined and new chunks were generated the cpu load runs constantly on 50% even if i leave the server. Only a complete stop/restart of it get it back down to idle. Than i join, fly to new chunks and same issue again.
I tried the recent SpongeForge versions and 1.11.2-2227-6.0.0-BETA-2171 runs correct, after 1.11.2-2227-6.0.0-BETA-2173 there is this issue. There seems to be something wrong with the commits made on 2173.
Does someone else notices a similiar thing or am i alone with this? i tried it on 2 different PC’s and it was the same.
Hope you guys could help, sponge is a great thing.
best regards
Flo
Thanks, could you post link to the thread you mentioned? i use the build 2171 for now. I also thank the spongy staff for their great work!
Only good thing on the new build was that the Dynmap generated better. I assume that sponge doesn’t support “chunkgenerate” as trigger for the dynmap.
No as i said in the main post i tried a clean server with no mods/config/data at all and the issue is the same. i better will start a new thread in the plugins section for my dynmap problem
Thank you really much for the quick update, but i have to disappoint you. This does not fix the issue completly.
If new chunks are generated something prevents the server to go back to idle.
It now doesn’t run on 50% CPU load, but on 10%. Normally when nobody is on the Server it runs between 1-10% and not constantly on 10% or more.
It’s far better but there still is an issue, you can try it yourself by running 2171 vs 2209.
I installed YourKit, If you say what exactly you need to know i could deliver a screenshot.
It is now running since more than a hour and back to full cpu load (50-80%) and doesn’t go back by itself also it gives me the “Can’t keep up” Message every few minutes, this is no problem on 2171 (1.11.2-2227-6.0.0-BETA-2171). There definitly still is something wrong.
Hey again, i don’t know if this helps any further. I think to get a better view on this i need to let it run for a longer time and do some things for example generate new chunks etc.
My Server normally runs on a Pentium G3440, i’m currently testing on my PC with a I5-3570K.
This is necessary that you understand that the current 10% CPU Load is not “good” in total idle, i’m on the Server doing exactly nothing. Just floating in the air. With the old version 2171 it idles correctly.
Edit: I can even leave the Server but the load persists.
Edit2: As longer the Server is running, the Load gets higher. It’s now up to 16% not going back down.
Unfortunately, I’m still unable to reproduce this - my YourKit threads view shows very little activity by the server thread, and /sponge tps consistently shows a mean tick time of around 2ms.
Does this happen if you create a fresh world on your server?
The last test i posted above was done with a fresh world, fresh server and fresh configuration without mods/plugins. It occured when i discovered some new chunks, it gets higher as more chunks i fly through and doesn’t go back to idle even when i leave the server than, this is very strange. I think i was flying like 100-200 chunks.
It even showed the “Can’t keep up” Message on my i5 PC with 4G ram allocated to the Server.
If you need further testing i need to check it again this evening…
I’m sorry to bother you with this…
Good News Aaron, with the new cleanup/changes the cpu load seems to drop back down if new chunks were generated. Thank you so much for the work! I need to pay you a beer or something.
But I think i found something else, on the new versions the server fires an massive amount of blockupdate event triggers specially when discovering new chunks. This seems to have to do with the light thing, but i’m not sure…
Good thing is that the dynmap now “works” because every chunk has so many blockupdate triggers that it renders them all - not as intended but thats a workaround for the missing chunkgenerate triggers. I think the fault here is not yours the Dynmap Developer seems to has not much time for his project anymore.
It has to sift through quite a bit of data.
