Nomploy Nomad Packs

← All packs

ollama v0.1.0

AI

Ollama — run open large language models (Llama, Mistral, Gemma, Qwen, and more) locally behind a simple REST API. Deployed as a host-networked Nomad service with a persistent Docker volume for downloaded models.

nomad-pack run ollama --registry nomploy

Source ↗ Project ↗ ★ 181.5k

1 task http 11434 1 volume image ollama/ollama:latest
Pairs with open-webuiqdrantweaviate
Variables 8
values.hcl

Save as values.hcl, edit, then run:

nomad-pack run ollama -f values.hcl --registry nomploy
# The name of the Nomad job.
job_name = "ollama"

# The Nomad namespace to deploy into.
namespace = "default"

# The datacenters to deploy to.
datacenters = ["*"]

# The Ollama container image. Pin a tag in production.
image = "ollama/ollama:latest"

# Host port for the Ollama REST API.
port = 11434

# Docker named volume for /root/.ollama (downloaded models — these can be many GB). Ollama runs as root, so a fresh volume is writable. Back it up (or just re-pull models).
data_volume = "ollama_data"

# Placement constraints — pin to one node so the local model volume stays put (and so you land on a node with enough RAM/CPU or a GPU). On a nomploy cluster: attribute = "$${meta.nomploy_control_plane}", operator = "=", value = "true".
constraints = []

# The task resources. LLMs are hungry — raise cpu/memory substantially for real models (7B+ needs several GB of RAM on CPU).
resources = {
    cpu    = 2000
    memory = 4096
  }
NameTypeDefaultDescription
job_name string 
"ollama"
 The name of the Nomad job.
namespace string 
"default"
 The Nomad namespace to deploy into.
datacenters list 
["*"]
 The datacenters to deploy to.
image string 
"ollama/ollama:latest"
 The Ollama container image. Pin a tag in production.
port number 
11434
 Host port for the Ollama REST API.
data_volume string 
"ollama_data"
 Docker named volume for /root/.ollama (downloaded models — these can be many GB). Ollama runs as root, so a fresh volume is writable. Back it up (or just re-pull models).
constraints list 
[]
 Placement constraints — pin to one node so the local model volume stays put (and so you land on a node with enough RAM/CPU or a GPU). On a nomploy cluster: attribute = "$${meta.nomploy_control_plane}", operator = "=", value = "true".
resources object 
{
    cpu    = 2000
    memory = 4096
  }
 The task resources. LLMs are hungry — raise cpu/memory substantially for real models (7B+ needs several GB of RAM on CPU).
Readme

ollama

Ollama — run open large language models (Llama, Mistral, Gemma, Qwen, DeepSeek, …) locally behind a simple REST API. Host-networked Nomad service with a persistent volume for downloaded models.

Usage

nomad-pack registry add nomploy github.com/Nomploy/nomad-packs
nomad-pack run ollama --registry nomploy

In nomploy: create a Compose service, type Nomad Pack, pack ollama, custom registry github.com/Nomploy/nomad-packs, then Deploy.

curl http://<node-ip>:11434/api/pull    -d '{"name":"llama3.2"}'
curl http://<node-ip>:11434/api/generate -d '{"model":"llama3.2","prompt":"hi"}'

Key variables

Variable Default Notes
image ollama/ollama:latest Pin a tag in production.
port 11434 REST API (OLLAMA_HOST).
data_volume ollama_data Downloaded models (can be many GB).
constraints [] Pin to a node with enough RAM (or a GPU) and to keep the model volume.
resources cpu 2000 / mem 4096 Raise substantially for real models — a 7B model needs several GB of RAM on CPU.

Notes