You can not select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
2.3 KiB
2.3 KiB
System Requirements — Ollama
Project: Ollama/
Tipe: infra (AI inference)
Last updated: 2026-07-06
Ringkasan
Ollama native di Ubuntu LTS — tanpa Docker. Inference CPU (tanpa GPU wajib).
Platform & OS
| Komponen | Minimum | Direkomendasikan | Saat ini (.26) |
|---|---|---|---|
| OS server | Ubuntu 22.04 LTS | Ubuntu Server 22.04 | 22.04.5 Desktop ⚠️ |
| Arsitektur | x86_64 + AVX2 | x86_64 | ✅ AVX2 |
| vCPU | 16 | 16–48 | ✅ 48 |
| RAM | 32 GiB | 64 GiB | ✅ 62 GiB |
| Disk model | 50 GB free | ≥ 100 GB di /data/ollama |
✅ 457 GB free |
Runtime & Toolchain
| Tool | Versi | Cek |
|---|---|---|
| Ollama | 0.30.8+ | ollama --version |
| systemd | — | systemctl status ollama |
| ufw | — | ufw status |
| curl | — | smoke test API |
Tidak dipakai: Docker, GPU driver, nginx di host ini.
Model (production CPU)
| Model | Peran | RAM load (approx) |
|---|---|---|
qwen2.5:7b |
Chat / answer | ~5–8 GB |
nomic-embed-text |
Embedding | ~1 GB |
Dependency eksternal
| Dependency | Host | Dipakai untuk | Owner docs |
|---|---|---|---|
| PostgreSQL + pgvector | 10.100.1.25 |
Vector index | server-connection |
| QNAP SMB | 10.100.1.10 |
Indexer source | server-connection |
| geonet-console | 10.100.1.24 |
API consumer | server-connection |
Variabel URL & konfigurasi
Dari .env.example:
| Variabel | Contoh | Keterangan |
|---|---|---|
OLLAMA_BASE_URL |
http://10.100.1.26:11434 |
Base URL API (consumer) |
OLLAMA_CHAT_MODEL |
qwen2.5:7b |
Model chat |
OLLAMA_EMBED_MODEL |
nomic-embed-text |
Model embedding |
OLLAMA_TIMEOUT |
120 |
Timeout detik (HTTP client) |
Server-side (systemd, bukan di repo):
| Variabel | Nilai live |
|---|---|
OLLAMA_HOST |
0.0.0.0:11434 |
OLLAMA_MODELS |
/data/ollama/models |
OLLAMA_NUM_PARALLEL |
2 |
OLLAMA_MAX_LOADED_MODELS |
2 |
Aturan: no hardcode URL/IP di kode aplikasi — gunakan env.
Network
- Ollama API hanya dari subnet
10.100.1.0/24 - Tidak perlu TLS internal (VLAN trusted); TLS di edge app jika perlu