Azerothcore-wotlk: Huge Memory Leak 6 hours = 1.694gb-> 2.810gb

[andoys]

Huge Memory Leak 6 hours = 1.694 -> 2.810gb

SMALL DESCRIPTION:

Huge Memory Leak 6 hours = 1.694 -> 2.810gb
check attach picture for more info

EXPECTED BLIZZLIKE BEHAVIOUR:

There should be no memory leak?

CURRENT BEHAVIOUR:

memory leak 6 hours = 1.694gb -> 2.810gb

STEPS TO REPRODUCE THE PROBLEM:

1. start server with 20+ people
2. wait 6-7 hours
3. just keeps on leaking

EXTRA NOTES:

BRANCH(ES):

master

AC HASH/COMMIT:

Azerothcore rev. ded199dd7e7b+

OPERATING SYSTEM:

Ubuntu 16/18

MODULES:

OTHER CUSTOMIZATIONS:

[BarbzYHOOL]

if it's a memory leak, then it will run out of memory and mega crash

never happened before

[Selyria]

So my server has been running for 12 days and I have a consumption of about 1.9GB ram on Windows 10 players are between 12-43 on it

and 2.8Gb I also find perfectly okay who there are not 10 gb is but everything in the green area 4GB are eh bissel little I find.

[BarbzYHOOL]

http://www.azerothcore.org/wiki/FAQ

[FrancescoBorzi]

In computer science, a memory leak is a type of resource leak that occurs when a computer program incorrectly manages memory allocations[1] in such a way that memory which is no longer needed is not released.

https://en.wikipedia.org/wiki/Memory_leak

You're only showing that the memory used by AC increases after a while. I don't see any leak here.

I'm closing this because it's not about a memory leak and not even about an issue. Thank you anyway for your contribution.

[fmrmx]

In fact there is a memory leak in AzerothCore, since the owners made a mess with grids and wintergrasp somewhere between 2017-2018, i fixed that long ago but i do not remember where the error was, but is related to gameobject active in grids.

i do not know why they keep closing issues, ah yea is the easy way.

[BarbzYHOOL]

not as easy as spitting bullshit

[FrancescoBorzi]

i fixed that long ago but i do not remember where the error was

so why didn't you share the solution by then? Or even just the steps to reproduce the problem, since I have never been able to do so in my local AC.

[poszer]

In fact there is a memory leak in AzerothCore, since the owners made a mess with grids and wintergrasp somewhere between 2017-2018, i fixed that long ago but i do not remember where the error was, but is related to gameobject active in grids.

i do not know why they keep closing issues, ah yea is the easy way.

Proofs? Way to reproduce it? Ah yes, there is no way

[fmrmx]

i am not interested in posting proofs, or ways to reproduce it, and i dont need to, i am not a server owner anymore, that was long ago, but indeed there are memory errors, stop trying to act like you all dont know, and closing issues, because there are like 10 reports of memory leaks and all closed, at least accept it lol.

[poszer]

i am not interested in posting proofs, or ways to reproduce it, and i dont need to, i am not a server owner anymore, that was long ago, but indeed there are memory errors, stop trying to act like you all dont know, and closing issues, because there are like 10 reports of memory leaks and all closed, at least accept it lol.

If there is, we need to find each memory leak source. The issue is accepted, but not reproduced :) If the Google Chrome use 2,5Gb than what's the problem of AC using 2,8 with some players online. Have a good day

[FrancescoBorzi]

@fmrmx it's not about accepting it or not. We close duplicate reports because it makes no sense to report the same (generic) thing multiple times (and it just creates caos). We demand details because without them we are not able to fix the memory issue(s), since most of the reports unfortunately are too generic.

[milestorme]

10 days uptime on my test for this memory hasn’t moved above 15% , I have grids set to not unload , I see no memory leaks at all I see very stable numbers

[fmrmx]

well at least check out some parts of the wintegrasp code , because there are some problems with AC, and yes 2.8g is a problem because most cores uses 1.3 to 1.6 gb of ram, and scales according to population, but AC increases up to 4-5gb i have observed that when some player logs in, there is a tiny memory increase, but when the player logs out, that tiny space remains forever until server restart, i am unable to reproduce the problem, because there is no glitch, it is simply memory usage increase, if it were an instance/spell/script problem, it would be more easier. i will atach you a screen of my core, so you can see the normal memory usage of a TrinityCore/SunwellCore based server.

2 days, 5 hours uptime.

i have

worldserver
authserver
mysql
apache
node
php

all running on the same machine, 2gb ram ubuntu 16.0x no problems at all

@milestrome , for low resource VPS/Dedicated, it is indeed a problem. you must have 10gb or more.
with grid unload disabled, it is normal to have 3-5gb but this also happen with gridunload activated.

[milestorme]

no memory leak here many months of testing, came from TrinityCore before Azerothcore, no memory leaks their either, this is paranoia without a shred of networking evidence , im guessing like most others with abnormal issues you modified your core, because AC runs very well especially the way it utilizes multi cores, but you all type top not top -H -p [pid] , geez , you stinged on hardware and are now screaming memory leak.

[fmrmx]

no memory leak here many months of testing, came from TrinityCore before Azerothcore, no memory leaks their either, this is paranoia without a shred of networking evidence , im guessing like most others with abnormal issues you modified your core, because AC runs very well especially the way it utilizes multi cores, but you all type top not top -H -p [pid] , geez , you stinged on hardware and are now screaming memory leak.

Not quite, what is the point of reporting issues of a modified core?, that is nonsense. lol, my hardware is perfectly working and i have advanced linux management knowledge, i dont stinged anything lol. you are not talking to a sKid dude.

[fmrmx]

ok i found a problem in this lines.

https://github.com/TrinityCore/TrinityCore/pull/7992/commits/cefcab32bbad8d4eadcc8d072da1932bcc4e23a3

for every user online, gameobjects loads, and stay in memory, if this two lines are deleted, it is possible to come back to 1.0x AzerothCore commits but that lines is a hack, TC never merged those changes, because of this:

https://github.com/TrinityCore/TrinityCore/pull/7992#issuecomment-9329785

so erasing that lines will possibly fix the memory usage problem. but i do not have the tools for testing it. but that lines was a hack, never merged into TC, due to bad memory management.

As always, before merging changes, it must be tested, because deleting this lines may break some wintergrasp functionality.

[FrancescoBorzi]

i have observed that when some player logs in, there is a tiny memory increase, but when the player logs out, that tiny space remains forever until server restart

Look, I haven't investigated the issue myself, but this is a good example of what should not taken for granted as a memory leak. It could be, for example, that the core stores this data about the player in the memory on purpose and then it keeps it there so when the player comes back the core doesn't have to reload the data from the database (which would instead be slower). Resulting in a overall performance increase at the price of more memory usage (that is fine since we have plenty of memory nowadays).

Of course, without an actual investigation we cannot exclude that it could just be that this memory is kept for no reason and it's just a programming mistake, thus a memory leak that should be indeed fixed.

What I'm trying to say here is just don't take for guaranteed that there is a problem when you see more memory used. Do investigate the code and the behaviour of the core before coming to a conclusion (being positive or negative).

[FrancescoBorzi]

I don't know what you mean by "storing in the memory" but whatever is not in the database will be lost at server reload.

then it is not correct to store every player data until relog

i have observed that when some player logs in, there is a tiny memory increase, but when the player logs out, that tiny space remains forever until server restart

If you log in and out the same character does the memory keep increasing? In that case there would be an issue for sure, otherwise we cannot determine whether it's a memory leak or not without additional context.

[FrancescoBorzi]

Also, just to remove any doubts (not sure if you meant that or not), this comment is clearly referring to "Azeroth" as the continent and not AzerothCore since the comment is from 2012 and we created AzerothCore only in mid 2016

[fmrmx]

Nevermind guys, i will not bother anymore. good luck.

[FrancescoBorzi]

@fmrmx no bothering at all, any technical discussion is always welcome

[fmrmx]

AzerothCore is based in SunwellCore, wich is based in TrinityCore, the lines i found, are found on your core too, that lines was a "hack" never merged to TC because of bad memory management, azeroth is the "world" of course. someone in the past merged that "hack" to your repository.

[FrancescoBorzi]

@fmrmx sure. I've also noticed that we have those lines while TC hasn't. Still we couldn't reproduce the issue mentioned in that discussion which sounds quite critical:

so with roundabout 50 players ur ram will start to burn

But still, if you have investigated it and think that those lines should be removed, it would be nice if you could come up with a PR that implements the needed changes and explain what would be impacted by them so we can test it accordingly.

[fmrmx]

@fmrmx sure. I've also noticed that we have those lines while TC hasn't. Still we couldn't reproduce the issue mentioned in that discussion which sounds quite critical:

so with roundabout 50 players ur ram will start to burn

But still, if you have investigated it and think that those lines should be removed, it would be nice if you could come up with a PR that implements the needed changes and explain what would be impacted by them so we can test it accordingly.

I am too lazy to do that. really lazy, especially now. well i am too lazy to write...

[Selyria]

@fmrmx sure. I've also noticed that we have those lines while TC hasn't. Still we couldn't reproduce the issue mentioned in that discussion which sounds quite critical:

so with roundabout 50 players ur ram will start to burn

But still, if you have investigated it and think that those lines should be removed, it would be nice if you could come up with a PR that implements the needed changes and explain what would be impacted by them so we can test it accordingly.

I am too lazy to do that. really lazy, especially now. well i am too lazy to write...

oh, man, this is supposed to be a joke, right?

They are too lazy to fix it but they can write very well see above.

[fmrmx]

@fmrmx sure. I've also noticed that we have those lines while TC hasn't. Still we couldn't reproduce the issue mentioned in that discussion which sounds quite critical:

so with roundabout 50 players ur ram will start to burn

But still, if you have investigated it and think that those lines should be removed, it would be nice if you could come up with a PR that implements the needed changes and explain what would be impacted by them so we can test it accordingly.

I am too lazy to do that. really lazy, especially now. well i am too lazy to write...

oh, man, this is supposed to be a joke, right?

They are too lazy to fix it but they can write very well see above.

it is your cores, not mine. as i said before, i am happy with my private core, but there are to many issues reporting this. if the core devs do not fix it, it is perfectly ok for me.

[FrancescoBorzi]

I hope you wrote your "private core" from scratch instead of violating the GPL license of an existing project.