[BUG] socket-proxy crashes dockerd

[andrebrait]

Is there an existing issue for this?

• I have searched the existing issues

Current Behavior

After running socket-proxy for a while, I notice dockerd dies after a while. I do not notice much else, other than some other operations with docker-compose and so on ending up in a segfault, but I think that's their CLI spinner thingy...

Some containers keep running, bit trying to eun dockerd nets the usual "cannot connect to socket, is docker running?" message.

If I manually start dockerd again, I usually see messages about being unable to mount/dismount images and something about a pivot_root.

This started happening somewhat recently. Previously I had run this container for months and nothing bad happened. Now I cannot deploy it for a full day without running into this issue.

Not running this container and directly mounting the docker socket file onto Traefik and Authentik works fine.

I have deleted my entire docker folder and rebuilt everything from scratch a number of times, but it always comes back except if, and only if, I do not run this container.

Expected Behavior

Ir keeps working, like it used to.

Steps To Reproduce

1. Deploy this container and have Traefik listen to it
2. Wait?
3. Crash

Environment

- OS: unRAID 7.2.4
- How docker service was installed: bundled with unRAID, version 27.5.1

Docker creation

# This is for Traefik

  dockerproxy:
    image: lscr.io/linuxserver/socket-proxy:latest
    container_name: traefik_socket-proxy
    restart: unless-stopped
    environment:
      CONTAINERS: 1
      POST: 0
    volumes:
      - /var/run/docker.sock:/var/run/docker.sock:ro
    read_only: true
    tmpfs:
      - /run


# And this is for Authentik

  dockerproxy:
    image: lscr.io/linuxserver/socket-proxy:latest
    container_name: authentik_socket-proxy
    restart: unless-stopped
    environment:
      CONTAINERS: 1
      POST: 1
      IMAGES: 1
      INFO: 1
    volumes:
      - /var/run/docker.sock:/var/run/docker.sock:ro
    read_only: true
    tmpfs:
      - /run

Container logs

I could not get it at the time, since docker was dead and I had to get my system back into operation.

I did see it consistently crash about once a day, every time, no matter what I did.

More details here (and possibly syslog, logs from docker itself and perhaps even containers)

https://forums.unraid.net/topic/197557-docker-crashing-docker-failed-to-initialize-socket-not-available-but-some-containers-keep-running-and-other-oddities

[github-actions]

Thanks for opening your first issue here! Be sure to follow the relevant issue templates, or risk having this issue marked as invalid.

[thespad]

So, one thing of note is that docker 27 is end of life and unsupported. That said it shouldn't fundamentally cause a problem.

I would suggest limiting logging on the container, as that can someone cause problems on busy hosts, I use something like

    logging:
      driver: json-file
      options:
        max-size: 100m
        max-file: "5"

Beyond that it's hard to toubleshoot without any kind of logs or errors and unfortunately the unraid forums do not allow guest users to access attachments.

[andrebrait]

I'll post the attachments here, but my server is rarely busy. I had the crash happen with just my Traefik stack (Traefik + Crowdsec + Redis + socket-proxy) + Plex (also linuxserver.io image) up, but no users were using anything.

As for logs from this container itself, I only remember seeing the occasional logs from it being queried by Traefik from time to time or something.

The crash also occurs within hours/days, whereas it ran just fine for months, previously.

My Docker folder is mounted onto a 4TB btrfs volume which had less than 50% utilization.

I'll try to enable logging like you suggested and seeing if I can get it from the docker folder when it happens.

nas-diagnostics-20260304-0629.zip
nas-diagnostics-20260304-0805.zip
nas-diagnostics-20260311-0206.zip