Infrastructure: What are the current Hetzner server roles and routing rules?

Documentation for Sessions

📝 Infrastructure: What are the current Hetzner server roles and routing rules?

Date: 2026-03-23
Time: 17:06:04 UTC
Session: gspace1-261.4
Agent: backend
Confidence: 95%
Status: ✅ Documented


📋 Decision Context

Question:

What are the current Hetzner server roles and routing rules?

Task: S261: Document server fleet roles for future agents


🎯 Decision Made

Chosen:

Two servers as of S261: (1) ax102-01 (AX102-U, 16C/32T Ryzen 9, 128GB, ax102-01.gograb.space) — roles: customer-sandbox, git-operations, priority 1, 60 VMs max. (2) ax42-01 (AX42-U, 8C/16T Ryzen 7, 128GB, sandbox.gograb.space) — roles: executive-sandbox, customer-sandbox (overflow), priority 2, 30 VMs max. Customer VMs route to ax102-01 first; if at capacity, overflow to ax42-01. Executive VMs route to ax42-01 first. Git operations (PLATFORM_GIT_SERVER_URL) route to ax102-01. The frontend proxy (/sandbox-api) also routes to ax102-01.

Options Considered:

  1. Single primary with hot standby (wasteful)
  2. Equal priority round-robin (no role optimization)
  3. Chosen: Role-based with priority and overflow

💭 Reasoning

AX102-U has 2x cores and 3D V-Cache — ideal for CPU-intensive Firecracker VMs and git operations. AX42-U is smaller but stable (23+ days uptime). Role separation ensures both servers are active concurrently, not primary/standby. Overflow routing means no wasted capacity. API keys are separate per server for blast radius containment.


📁 Files Involved


🔗 Related


Auto-generated by MCP Smart Consolidation