← Back to DevOps Fixes

Docker Exit Code 137: How to Diagnose and Fix OOM-Killed Containers

1. Confirm OOM Kill in 60 Seconds

Run these commands on your host node or container daemon to verify if OOM was the cause:

docker inspect <container_name_or_id> --format '{{.State.OOMKilled}}'

If the output is true, the cgroup limit was breached. You can also inspect the memory allocation limit configured for the container:

docker inspect <container_name_or_id> --format '{{json .HostConfig.Memory}}'

2. Common Causes of Exit Code 137

1. Unbounded Runtime / JVM Heap: Java applications running inside Docker without container-aware heap flags (e.g. -XX:MaxRAMPercentage=75.0) often allocate memory based on total host RAM rather than the container cgroup limit.
2. Memory Leaks in Background Workers: Unclosed database connection pools or memory buffering in Node.js/Python workers gradually grow until the cgroup limit is hit.
3. Node Memory Starvation: If container memory limit is set to 0 (unlimited), the kernel OOM killer terminates containers when the physical host node runs out of RAM.

3. Step-by-Step Resolution

To resolve Exit Code 137 permanently:

• Increase container memory limits in your docker-compose.yml or Kubernetes deployment manifest:

  services:
    app:
      image: myapp:latest
      deploy:
        resources:
          limits:
            memory: 2G
          reservations:
            memory: 1G

• If running memory-intensive workloads like build runners or analytics, consider running them on dedicated cloud nodes with adequate RAM. Spin up high-memory cloud droplets on DigitalOcean to isolate build pipelines.
• For hardware upgrades, check out recommended server RAM and hardware on Amazon using tag dev0ps-20: ECC Server Memory Deals on Amazon (tag dev0ps-20).
• Recommended reading on high-concurrency architecture: Designing Data-Intensive Applications on Amazon.