i2p front end inaccessible after its container is restarted - eviltoast

I’m experimenting with i2p and a librewolf container setup in Docker compose. However, the i2p web console front end (127.0.0.1:7657) becomes inaccessible if the container itself is restarted. This can be remedied by removing the directories that get created by the volume mappings in the compose file, but this obviously not ideal. Anyone have experience with this problem? I’ve seen hints from people online suggesting that the data in those directories getting somehow corrupted. I have not yet investigated that further.

version: "3.5"
services:
  i2p_router:
    image:
      geti2p/i2p:latest
    environment:
    - JVM_XMX=256m
    volumes:
    - ./i2phome:/i2p/.i2p
    - ./i2ptorrents:/i2psnark
    ports:
    - 4444:4444
    - 6668:6668
    - 7657:7657
    - 9001:12345
    - 9002:12345/udp

  libre_wolf:
    image:
      linuxserver/librewolf
    ports:
    - 9300:3000
    - 9301:3001

volumes:
  i2phome:
  i2ptorrents:
networks:
  frontend:
    driver: bridge
  • NegativeLookBehind@lemmy.worldOPEnglish
    1·
    2 months ago

    Doing more digging, I captured checksums of all files that get created by the i2p container, with

    find i2p* -type f -exec md5 {} + > 1sum # After starting container

    and

    find i2p* -type f -exec md5 {} + > 2sum # After restarting container

    Comparing the files with diff -y 1sum 2sum, I noticed this file was the only one whose checksum changed:

    MD5 (i2phome/router.ping) = 95fd0fa14b084cf019f2dd9e0884aa1 | MD5 (i2phome/router.ping) = f4b060ae4f789f7d24f2851c06597c4

    EDIT:

    Preserving this file outside of the volume directory and copying it back in after the container restarts allows the web console to function as expected.

    EDIT 2:

    To take it a step further, I’m mounting it read only into the container at runtime like this:

    - ./router.ping:/i2p/.i2p/:ro

    which is sort of a hack, but I think I can live with it.