Alejandro Bentivengo

I think its time for the mojang team to take over this issue. This is a very serious problem, if the app is leaking memory, its stability is at risk.

I've already said that if needed I can provide access to a remote linux console to the server, and any access needed to investigate the issue.

For the mod: please take this as a real problem, there is already enough evidence in this ticket to move to confirm and its still not happening. Why is this it?

the description of the problem seems to fail to mention what you are saying.

For what I can understand from it, its the same problem I'm having with my server since 1.8 came out.

And again, we are not talking about an "increase of memory" we are talking of a hungy beast memory server which uses and colapses the system after consuming all its resources. Im not a computer expert, but I've studied and researched memory usage, the system takes after no longer needed its released. Its a cycle... which is not being done with the minecraft servers since 1.8 (prereleases).

If there is new information about the ticket that might point this problem in another direction it would be nice to have it in the description.

Kumasasa, running the server with the default Arguments (published with the server JAR file in the minecraft.net site) gives the same results.

I've spent hours trying different parameters, an ALL of them give the same result: heap usage increases, but NEVER decreases.

@Ryan I have a MineOS server which I belive is Debian based with kernel 3.x
uname -r
3.2.0-4-amd64

I can confirm that this problem is hapenning to me, not sure if thats what u wanted to check.

I've restarted the server 45 minutes ago aprox.
This is the status right now... more than half of the memory used already, and none being returned to the system.

I've tried that, memory filled up in less than half an hour.
I will post my server status with 3 players online right now in a sec.

I just checked my server. After running for about 10hs, with a maximum of 3 persons online and 1 currently, it stands at 4500mb out of 5000mb used.
Something nasty is definetly going on. It just cant be possible that a server keep draining memory without releasing at least some.

Also, i checked the entities IDs, with fresh start, the entities IDs (seen on the log when someone joins the server) are usually pretty low. Between 500-1000. But if you check after some hours, the entities IDs rise to 100.000 and even more.

Are servers not unloading and disposing entities correctly?

I've never allowed the server to run at max mem usage just because it MIGHT crash and corrupt the world save. So I have no idea what happens.

After testing the oracle version, everything seems exactly the same. The memory builds up but its never recovered.

This might also be an object pooling system being implemente into the game, but its extremely strange that a server starts using memory unit it reaches 100% without releasing some of the unused back to the system.

Of couse im saying this taking into account that entitied shouldnt move before the chunk is loaded. Otherwise you would find dead chicken, cows, even villagers everywhere sofocating in walls and stuff. Again, unless transparent blocks are not even seen by entities, in that case they could just traverse them. But, right here is where I at least hit the wall as im just talking off without even knowing the source code. We could be right or we could be as wrong as it can get, hopefully mojang fixes this soon. As far as I've heard, in the last patch overhaul, they did some intresting patching to the block data itself, so might finally be seeing the light at the end of the tunnel here 😛

@CaptainStarbuck i dont think thats the solution to the problem. 1 + 2 = 3, as much as 2 + 1 = 3. The way you unload the chunks and entities doesnt really matter imo. A valid solution could be that entities inside a transparent block as fences al teleported to an adyacent free air space. If we say the entitie is at X: 3.4, Y: 70, Z: 3.47, it should be teleported to X: 3, Y: 70, Z: 3. This would be just a quick fix that i can think of as im writting this. No idea of its aplicable or not. What Im trying to say though is that it doesnt matter the order in which you unload stuff. Or at least it shouldnt 😛

I've just updated my java version from OpenJDK to the last Oracle Java JRE (1.7.0_67)

Will come back with results in a few hours.

it fills up until it reaches the maximum memory allocated for the server (java_xmx)...

I've tried different java tweeks to try to reduce the problem, but its certenly not a problem with the JVM. This problem started with this last version of the game, on 1.7 i didnt have any problems at all.

I will publish down here the setting im using right now for the server, after trying many many (MAAAAANY) different tweeks, this onces worked the best by keeping a relatively low ammount of memory used:
-XX:+UseConcMarkSweepGC -XX:+UseParNewGC -XX:+CMSIncrementalPacing -XX:ParallelGCThreads=2 -XX:+AggressiveOpts -Xincgc -XX:PermSize=512m

This is how my server looks right now, its been running for almost 12-14hs. Now there is NO ONE connected, but its still using 3.3GB of memory. What for? I have no idea... It just keeps taking memory, but never gives it back. Looks like somewhere chunks are just loaded and loaded, but never cleaned from the memory after usage... or maybe entities? or maybe both? No idea...

Uploading image with my current server running a little over 12hs with no players connected. Terrible memory leak...

Please check my thread MC-67517 as I believe this is a duplicate. I am willing to let mods log in remotely if necesary to fix the problem.

I can confirm this happens to me too on a MineOS VM

BTW, this I believe is the same report as MC-67517

Owner, please check if it relates to that problem.

I think you are right Steven, I just checked my server log and it looks like it cant keep up. Its so strange though... like 3Gb RAM and 4 cores working at 3.0Ghz should be powerfull enough to handle this... I'm attaching the server log ASAP..

Its strange that with older versions (1.6.4) I could get over 1000 entities and with newer versions (I only tested this on the last snapshots) it doesnt work...

I've just increased the amount of threads for the VM and the amount of RAM assigned to 5Gb. I'll try again and come back with the results

Check the system specs, there is no reason for a server to crash with only 500 entities. I could host a single player map and get more than 1000 with half the ram assigned to the game.
This has nothing to do with server capacity, as the server works perfectly without any lag seconds before it crashes.
If it was a problem with the capacity of the server, lag would be extremly noticeable much before it crashes.

Sheep are being affected for me. Today I found 2 of my blue sheep roaming around close by outside of their pen.

Confirmed on 1.8pre-1

still happening in last snapshot (14w34b) on mutliplayer. Just logged off from my server and a couple of animales with name tags were roaming around out of the fenced areas.