Autopausing knocked in loop

[attssystem]

Hi :)

I'm using this docker image with this compose file :

name: minecraft
services:
  minecraft:
    image: itzg/minecraft-server:latest
    container_name: minecraft
    ports:
      - 25565:25565
    volumes:
      - ./data:/data
    environment:
      VERSION: "1.21"
      INIT_MEMORY: "512M"
      MAX_MEMORY: "2048M"
      EULA: "TRUE"
      ENABLE_AUTOPAUSE: "TRUE"
      MAX_TICK_TIME: "-1"
      # Aggressive for tests
      AUTOPAUSE_TIMEOUT_INIT: "10"
      AUTOPAUSE_TIMEOUT_EST: "1"
    restart: always

Autopause seems working at first glance but in fact it always get knocked :

minecraft  | [2024-08-10T14:04:13+00:00] [Autopause] Pausing Java process
minecraft  | [14:04:44] [Server thread/WARN]: Can't keep up! Is the server overloaded? Running 30946ms or 618 ticks behind
minecraft  | [Autopause loop] Server was knocked - waiting for clients or timeout
minecraft  | [Autopause loop] No client connected since startup / knocked - pausing
minecraft  | [14:06:48] [RCON Listener #1/INFO]: Thread RCON Client /0:0:0:0:0:0:0:1 started
minecraft  | [14:06:48] [Server thread/INFO]: [Rcon: Saved the game]
minecraft  | [14:06:48] [RCON Client /0:0:0:0:0:0:0:1 #12/INFO]: Thread RCON Client /0:0:0:0:0:0:0:1 shutting down
minecraft  | [2024-08-10T14:06:48+00:00] [Autopause] Pausing Java process
minecraft  | [14:07:40] [Server thread/WARN]: Can't keep up! Is the server overloaded? Running 52284ms or 1045 ticks behind
minecraft  | [Autopause loop] Server was knocked - waiting for clients or timeout
minecraft  | [Autopause loop] No client connected since startup / knocked - pausing
minecraft  | [14:09:43] [RCON Listener #1/INFO]: Thread RCON Client /0:0:0:0:0:0:0:1 started
minecraft  | [14:09:43] [Server thread/INFO]: [Rcon: Saved the game]
minecraft  | [14:09:43] [RCON Client /0:0:0:0:0:0:0:1 #13/INFO]: Thread RCON Client /0:0:0:0:0:0:0:1 shutting down
minecraft  | [2024-08-10T14:09:43+00:00] [Autopause] Pausing Java process

It looks like the java process isnt properly paused and raises warning because of overloading (kinda logic if the server is supposed to be paused).

By bashing into the container, the pausing mechanism put the java process in Tl state which seems ok and then the above warn message wakes the server up again and again.

I tried with longer delays, with JVM_DD_OPTS: "disable.watchdog=true" but nothing change. Also tried server 1.19.2 which worked a few month ago.

Do you hava any idea ?

[itzg]

Ignoring that warning, which you can and should do, it sounds like the software is correctly pausing the process and then later correctly processing an external network access to resume the server.

You'll need to investigate what is accessing the port. Perhaps use network tracing software like Wireshark.

[attssystem]

You're fast ! Thx ! I know this warning can easily be ignored but as both happen always at the same time I thought it could be linked. But in fact maybe server is knocked and started and always generate the warning. Orders of the message could be ignored.

I'll investigate myself then, thx again.

[Knufle]

I was about to make a post when I saw your discussion, I have the exact same issue, were you able to figure it out?

[attssystem]

Nothing more sorry. On top of that I didn't started it in a long time. I'm wondering, if it's facing the internet, if it's maybe just a lot of bots ? Seems too periodic tho