Nomad — Scheduler Without the Extra Baggage
What it is
Nomad is HashiCorp’s cluster scheduler. Its job is narrow: take a workload and decide where it should run. No complex layers, no endless plugins. It doesn’t try to be Kubernetes — and that’s the point. You drop in a job, Nomad finds a host, and the thing runs.
How it runs
The system is split into servers and clients. Servers keep state and make scheduling choices. Clients are the machines where jobs actually land. Jobs are described in HCL files — the same style HashiCorp uses across tools. Those jobs can be Docker containers, Java apps, simple binaries, even VMs in some setups. Nomad then allocates them to available nodes. If you need service discovery or secrets, you usually pair it with Consul and Vault.
Technical profile
Area | Details |
Role | Cluster scheduler for mixed workloads |
Platforms | Linux, Windows, macOS (as client) |
Workloads | Containers, binaries, JARs, VMs |
Control | CLI, REST API, web UI |
Cluster model | Multi-node, multi-region supported |
Add-ons | Consul (service discovery), Vault (secrets) |
Distribution | Single binary, minimal dependencies |
License | MPL 2.0, open source |
Setup notes
– Download the binary, put it in PATH.
– For local test: nomad agent -dev.
– For real clusters: run servers and clients with configs.
– Write a job spec in HCL.
– Launch with nomad run job.nomad.
– Check allocation and logs with nomad status.
Where it shows up
– Shops running both containers and legacy binaries side by side.
– Environments where Kubernetes feels like too much overhead.
– Multi-datacenter setups with simple scheduling needs.
– Labs and dev clusters that need quick job submission and cleanup.
Gaps to expect
– Service discovery and secrets are not built-in — must be paired with Consul/Vault.
– Smaller ecosystem compared to Kubernetes.
– Networking is basic; advanced cases require extra tools.
– Adoption mostly inside teams already using HashiCorp stack.
Quick comparison
Tool | What stands out | Fits best |
Nomad | Lightweight, runs anything, one binary | Mixed jobs, lean clusters |
Kubernetes | Full ecosystem, autoscaling | Enterprises with complex container needs |
Docker Swarm | Simple, Docker-native | Small container-only clusters |
Mesos/Marathon | Scale-out, multi-framework | Specialized, very large deployments |