Docker Ephemeral Volume Fix

Docker Ephemeral Directory Config Loss

Context: Docker containers with app migration/rebrand that writes active config to a NEW path not covered by existing volume mounts

Problem

When a containerized app rebrands (e.g., legacy-app → myapp), the migration creates a new config directory (.myapp/) alongside the old one (.legacy-app/). If docker-compose.yml only mounts the OLD directory, the new one is ephemeral — any changes made via dashboard/UI/API are lost on docker compose down && up.

Symptoms:

• Config changes (made via dashboard or API) revert after container recreation
• Settings "randomly" reset (actually on every down/up or recreate)
• Migration tool logs "copying config" on every startup
• Works fine after docker compose restart (no recreation) but breaks after down/up

Root Cause Analysis

docker-compose.yml:
  volumes:
    - /root/.legacy-app:/root/.legacy-app    # ← Mounted (persists)
    # /root/.myapp NOT mounted          # ← Ephemeral!

Startup flow:
1. Container created → .myapp/ is empty (ephemeral)
2. "Doctor" migration copies .legacy-app/ → .myapp/
3. App runs, dashboard writes to .myapp/ (active config)
4. docker compose down → container destroyed, .myapp/ GONE
5. docker compose up → back to step 1, dashboard changes lost

Solution

1. Copy live config from running container to host BEFORE destroying it:

   mkdir -p /root/.myapp
   docker cp <container>:/root/.myapp/. /root/.myapp/
   

2. Add volume mount to docker-compose.yml:

   volumes:
     - /root/.legacy-app:/root/.legacy-app
     - /root/.myapp:/root/.myapp    # NEW
   

3. Also copy nested auth/credential files that may live under subdirectories:

   # Check for auth files in old path that may not have been migrated
   find /root/.legacy-app -name 'auth-profiles*' -o -name '*.key' -o -name 'credentials*'
   # Copy any missing ones to the new path
   cp /root/.legacy-app/agents/main/agent/auth-profiles.json \
      /root/.myapp/agents/main/agent/auth-profiles.json
   

4. Recreate container: docker compose down && docker compose up -d

5. Verify migration skips: Look for log like "State dir migration skipped: target already exists"

Diagnostic Checklist

# 1. Check what's mounted
docker inspect <container> --format '{{range .Mounts}}{{.Source}} -> {{.Destination}} ({{.Type}})
{{end}}'

# 2. Check if app writes to an unmounted path
docker exec <container> find /root -name '*.json' -newer /proc/1/status -type f

# 3. Compare host vs container config
diff <(cat /host/path/config.json | jq -S .) \
     <(docker exec <container> cat /container/path/config.json | jq -S .)

When to Use

• Container app config keeps reverting after restarts
• Dashboard/UI changes don't persist
• App went through a rebrand/migration with new config paths
• docker compose restart works but down/up doesn't