Description

Repository

• https://github.com/NVIDIA/NemoClaw
• Branch: main

Description

Neither Dockerfile nor Dockerfile.base includes a HEALTHCHECK instruction. In standalone Docker deployments (without Kubernetes probes), Docker and Docker Compose cannot detect unhealthy containers. Containers may appear running but be internally broken.

Impact

Failed containers remain in "running" state, causing silent outages in non-K8s deployments.

Affected Area

• File(s): Dockerfile, Dockerfile.base

Expected Behavior

HEALTHCHECK instruction enables Docker to detect and report unhealthy containers.

Actual Behavior

No health monitoring available for standalone Docker deployments.

Reproduction Steps

Steps to Reproduce

1. docker build -t nemoclaw .
2. docker run -d nemoclaw
3. Kill the main process inside: container still shows "Up"
4. docker inspect --format='{{.State.Health}}' nemoclaw → no health data

Environment

• OS: Any (Docker host)
• NemoClaw Version: v0.1.0
• Branch: main
• Docker Version: 24.x+
• Runtime: Docker
• Container / Orchestration Info: Standalone Docker (not K8s)
• Network Setup: N/A

Debug Output

# Verify no HEALTHCHECK exists in the Dockerfile:
grep -i 'HEALTHCHECK' Dockerfile Dockerfile.base
# Expected: no matches

# Check container health status:
docker inspect --format='{{.State.Health}}' <container-id>
# Expected: <nil> (no health check configured)

# Manual health verification:
docker exec <container-id> curl -sf http://127.0.0.1:18789/health || echo "unhealthy"

Logs

# docker ps output showing no health status:
CONTAINER ID   IMAGE      STATUS          PORTS   NAMES
abc123         nemoclaw   Up 5 minutes            my-nemoclaw
# ↑ No "(healthy)" or "(unhealthy)" status — HEALTHCHECK not configured

Checklist

• I confirmed this bug is reproducible
• I searched existing issues and this is not a duplicate