Skip to content
Closed
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
25 changes: 25 additions & 0 deletions docs/design/b-0590-slice-5-os-choice.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,25 @@
# B-0590 Slice 5: OS Choice Options

## Objective
Evaluate and select the optimal OS choice for the 20-machine Otto fleet.

## Options

### 1. NixOS (Declarative-fleet)
- **Pros:** Completely declarative configuration. Excellent for maintaining identical state across 20 machines.
- **Cons:** Steeper learning curve, potentially harder for the swarm to auto-repair if derivation fails.

### 2. Debian/Ubuntu (Pragmatic)
- **Pros:** Standard, predictable. Easy to script via `bash` or `ansible`. Huge community support and agent familiarity.
- **Cons:** Imperative configuration can lead to configuration drift across the 20 nodes over time.

### 3. Talos (k8s-only)
- **Pros:** Minimal OS, API-driven, designed specifically for Kubernetes. Perfect if we are moving towards a K8s swarm architecture.
- **Cons:** Locks us entirely into the Kubernetes ecosystem, which may overcomplicate pure Otto KVM tasks.

### 4. Proxmox (VM-per-Otto)
- **Pros:** Excellent for virtualization. Allows segmenting Otto into discrete VMs and snapshotting perfectly.
- **Cons:** Heavier overhead, adds a hypervisor layer which may complicate pure bare-metal tasks and GPU passthrough (though OCuLink support exists).

## Recommendation Draft
Pending multi-agent review, **Debian/Ubuntu** offers the highest pragmatic velocity for immediate deployment, paired with rigorous state scripts to prevent drift. However, **Proxmox** remains a strong contender if we require rapid rollback via snapshots.
Loading