#Docker container ram usage
80 messages · Page 1 of 1 (latest)
That's how JVM GC works
Can i somehow optimize it? 😄
Try this https://flags.sh/
But the current described behavior looks like he's already using Aikar flags
Hm
Ok but
What do you mean by optimize
Send me your java flags
Using Aikar flags would improve performance a lot but it will also increase memory usage which tbh shouldn't be a reason to worry about
java -Xmx12288M -Xms12288M -XX:+UseG1GC -XX:+ParallelRefProcEnabled -XX:MaxGCPauseMillis=200 -XX:+UnlockExperimentalVMOptions -XX:+DisableExplicitGC -XX:G1NewSizePercent=40 -XX:G1MaxNewSizePercent=50 -XX:G1HeapRegionSize=16M -XX:G1ReservePercent=15 -XX:G1HeapWastePercent=5 -XX:G1MixedGCCountTarget=4 -XX:InitiatingHeapOccupancyPercent=20 -XX:G1MixedGCLiveThresholdPercent=80 -XX:G1RSetUpdatingPauseTimePercent=5 -XX:SurvivorRatio=32 -XX:+PerfDisableSharedMem -XX:MaxTenuringThreshold=1 -Djava.awt.headless=true -Daikars.new.flags=true -Dusing.aikars.flags=https://mcflags.emc.gs -jar skyblock.jar --nogui
40-50gb for day is a lot i think 😄
hm
For container only
You should just perform restarts every 3 AM or sth
That's unfortunately how JVM works
Memory usage will be growing and growing
I perform restart every day at 4am but 30 gigs increase in about 4-6 hours
Yes minecraft is running max 10gb but container 30 xd
The other cause of the growth may be memory leaks in your plugins
It shouldn't grow that fast
That should increase minecraft server ram not only container
Doesn't it increase Minecraft server RAM usage as well?
Nope, just container
Me too 😄 Its my first time with this problem
i have never had a problem like that

My container sometimes uses 37 GB but it grows up to it after 24 hours from starting the server with 32 GB flag
Can i somehow create heapdump from docker? 😄
Yes 😄
concurrently?
I think Ill soon run servers in screen xd
If you're setting your Xmx to 10GB and you're container memory is filling up to 30 GB, your server has a memory leak. Nothing do with pterodactyl/docker
And it's a leak in native memory space
Is it possible on clear paper server with 0 plugins?
If you have a memory leak, there is no way to "clear" it. You need to fix or remove what ever code/plugin is causing it
Yeah
And then you should report it to Paper
It seems like it's fault of docker/wings or something like that. If i start the same server in screen there is no problem with ram usage.
@mystic solstice I think it might be a bug with the most recent panel update. I'm getting this exact problem-- spigot reports proper ram usage, the panel reports 4GB+ more.
This is on the latest panel version, correct?
Yes
at worst, it's a docker issue. The panel doesn't really have any way to effect it.
but no issues on any of my minecraft or other Java servers
yeah, makes sense. i'll try restarting and such
I had the same issue but with other game
try to update docker and disable swap memory from panel
but is strange, the same docker version with wings 1.5.3 (panel 1.7.2) the issue is not visible
GRUB_CMDLINE_LINUX_DEFAULT="swapaccount=1 cgroup_enable=memory"
and put this on grub linux & linux default
GRUB_CMDLINE_LINUX=“swapaccount=1 cgroup_enable=memory"
I have the same issue, I made dumps and analyzed them to see if there is any leak, couldn't find anything.
Is there another way to find a possible leak without removing the jars one by one?
I have the same situation as the OP, +100 players in one server, 16gb ram for Minecraft and 32 for the container.
The ram fills up until it goes OOM, even with 102gb for the container OOM can happen
same here!
Yeah, when I use a backup on my Windows machine outside of wings and such, I don't get this issue. Very strange. I may set up a test ptero server on my Linux partition to test version as well, maybe I can reproduce it consistently.
where exactly would I add it? Is there a grub config for docker, or is the host system?
/etc/default/grub
Do you fix?
or someone have fixed?
i dont think its a bug
the JVM has some additional overhead thats not seen via the spigot readouts
(which is why aikar specifically called out to ptero users to not allocate the entire amount of RAM to the server in his optimization guide)
to allow for JVM overheads
wait, 18.4 for 11.7, weird
could you create a heap dump or smth similar using spark

