from Hacker News

Ask HN: Go libraries for managing Docker container pools and executing commands?

by magundu on 4/2/25, 6:31 PM with 3 comments

I’m developing a system in Go that maintains a fixed pool of Docker containers (e.g., 10) running a specific image (like ‘node’), where each container remains alive (using a command like tail -f) to be ready for executing arbitrary commands via docker exec. The system tracks the workload of each container, distributes commands to the least loaded one, and monitors container health to automatically restart or replace unhealthy instances.

I’m aware of the official Docker Go SDK (github.com/docker/docker/client) for managing containers, but I’m curious if there are any higher-level tools or libraries in Go that provide additional support for scheduling, load balancing, or enhanced health monitoring of containers in such a setup. Has anyone built or used libraries that streamline this kind of container orchestration and command execution?

Any insights, recommendations, or experiences would be greatly appreciated!

  • by SonuSitebot on 4/5/25, 1:15 PM

    I've worked on a similar setup in Go — managing a pool of "always-on" containers for isolated task execution via docker exec. The official Docker SDK is solid but pretty low-level, so I get the desire for something more ergonomic. In my experience, there aren't many off-the-shelf Go libraries that give you full orchestration primitives (load balancing, health checks, scheduling) out of the box like you'd find in Nomad or K8s. But here are a few options worth exploring:

    gofiber/fiber – not container-specific, but useful for building lightweight async schedulers if you're rolling your own orchestration logic.

    dockertest – primarily for testing, but you can adapt its logic for simplified lifecycle management.

    hashicorp/go-plugin – good for decoupling workloads, especially if you're considering container-based isolation per plugin/command.

    That said, most teams I’ve seen build their own lightweight layer on top of the Docker SDK with Redis or internal queues for tracking load/health. Curious if you're doing multi-host management or keeping this local? Also, make sure to aggressively timeout and clean up zombie exec sessions — they sneak up fast when you're doing docker exec a lot.

    Would love to hear more if you open source anything from this!