The Log Collector is the FleetManager’s black-box flight recorder: a small container that runs next to your workload and continuously pushes diagnostic data to the FleetManager — so root-cause information exists centrally before a machine goes offline.

It is a generic add-on: it runs on any Docker host (edge appliance, VM, bare metal), stays unprivileged, and talks only to a read-only docker-socket-proxy — never the raw Docker socket.

What it collects

• Tier 1 — host metrics: CPU, load, RAM, disk and container state, about every 90 seconds (offline-resilient, deduplicated).
• Tier 2 — log bundles: compressed tar.zst snapshots of Docker container logs + host diagnostics (df, ps, /proc, docker inspect), uploaded in resumable chunks. Logs are mirrored to a host volume continuously, so a bundle assembled after a crash still contains the pre-outage lines.
• On demand: the “Collect logs now” button in the dashboard pulls a fresh bundle.
• Event-triggered: a container crash/OOM fires an immediate bundle.

Installation

1. Fleet config

Create a config.json with your fleet identity (customer, machine and an ingest token fmt_… from the FleetManager):

{
  "fleetmanagement": {
    "base_url": "https://fleet.brinkhaus-gmbh.de",
    "customer_name": "your-customer-slug",
    "machine": "your-machine-name",
    "token": "fmt_YOUR_INGEST_TOKEN",
    "heartbeat_interval_sec": 60
  }
}

2. Docker Compose

services:
  docker-socket-proxy:
    image: tecnativa/docker-socket-proxy:0.3.0
    restart: always
    environment:
      CONTAINERS: 1
      INFO: 1
      EVENTS: 1
      POST: 0
    volumes:
      - /var/run/docker.sock:/var/run/docker.sock:ro

  logcollector:
    image: brinkhausgmbh/fleetmanager-logcollector:latest
    restart: always
    depends_on:
      - docker-socket-proxy
    environment:
      DOCKER_PROXY_URL: tcp://docker-socket-proxy:2375
      METRICS_DISK_PATH: /host-root
      SPOOL_DIR: /spool
    volumes:
      - ./config.json:/app/fleetManagementData/config.json:ro
      - logcollector-spool:/spool
      - /:/host-root:ro
    cpus: 0.50
    mem_limit: 256m

volumes:
  logcollector-spool:

3. Start

docker compose up -d
docker compose logs -f logcollector

Within a minute the machine appears in the dashboard — with a Resources tab (metrics) and an Artifacts tab (downloadable log bundles).

Configuration

Fleet identity lives in config.json. Operational behaviour is driven by environment variables — the most important:

Security

The collector never mounts the raw /var/run/docker.sock. All Docker access goes through the read-only docker-socket-proxy (CONTAINERS/INFO/EVENTS=1, POST=0). POST=0 blocks every state-changing Engine call. The container runs as an unprivileged user; the spool is bounded and can never fill the disk.