infrastructure

Infrastructure Management

Quick Reference

Deploy with Colmena

# Single host
colmena apply --on <hostname> --impure

# Multiple hosts
colmena apply --on host1,host2,host3 --impure

# Build only (no deploy)
colmena build --on <hostname> --impure

Proxmox Container Management

SSH to Proxmox host first, then use pct:

# List containers on a host
ssh <proxmox-host> "pct list"

# Container status
ssh <proxmox-host> "pct status <vmid>"
ssh <proxmox-host> "pct status <vmid> --verbose"

# Start/stop/reboot
ssh <proxmox-host> "pct start <vmid>"
ssh <proxmox-host> "pct stop <vmid>"
ssh <proxmox-host> "pct reboot <vmid>"

# Execute command in container
ssh <proxmox-host> "pct exec <vmid> -- /run/current-system/sw/bin/<command>"

# Common commands via pct exec
ssh <proxmox-host> "pct exec <vmid> -- /run/current-system/sw/bin/systemctl status <service>"
ssh <proxmox-host> "pct exec <vmid> -- /run/current-system/sw/bin/journalctl -u <service> -n 50"

Server Inventory

Proxmox Hosts

Special Hosts (NOT on Proxmox)

Larussa Post-Deploy Verification

After deploying to larussa, always verify GPU access:

# 1. Deploy to larussa
colmena apply --on larussa --impure

# 2. Verify NVIDIA GPU is accessible
ssh larussa nvidia-smi

If nvidia-smi errors out, the NVIDIA driver failed to reload properly. Reboot is required:

# Reboot larussa
ssh larussa sudo reboot

Recovery timeline:

• Machine reboot: ~3-4 minutes
• Podman containers start: additional ~1-2 minutes
• Total: ~5-6 minutes before fully operational

Verify recovery:

# Check if machine is back
ssh larussa hostname

# Verify GPU after reboot
ssh larussa nvidia-smi

# Check podman containers are running
ssh larussa podman ps

If issues persist, use SSH to diagnose:

ssh larussa journalctl -u nvidia-persistenced -n 50
ssh larussa systemctl status podman.socket
ssh larussa podman ps -a

Proxmox LXC Containers

All other hosts are LXC containers. Use pct list on Proxmox hosts to see VMIDs.

Common hosts: gitea, gitea-runner-1/2/3, prometheus, grafana, uptime-kuma, sonarqube, jellyseerr, prowlarr, n8n, minio, scanner, paperless, woodpecker, resume, glance, external-metrics

Troubleshooting Workflows

Container Won't Respond

1. Check status: ssh <proxmox-host> "pct status <vmid> --verbose"
2. If running but commands fail: ssh <proxmox-host> "pct reboot <vmid>"
3. Wait 15-30 seconds, verify: ssh <proxmox-host> "pct status <vmid>"
4. Re-deploy if needed: colmena apply --on <hostname> --impure

Service Not Working

1. Check service status:

   ssh <proxmox-host> "pct exec <vmid> -- /run/current-system/sw/bin/systemctl status <service>"
   

2. Check logs:

   ssh <proxmox-host> "pct exec <vmid> -- /run/current-system/sw/bin/journalctl -u <service> -n 100"
   

3. Restart service:

   ssh <proxmox-host> "pct exec <vmid> -- /run/current-system/sw/bin/systemctl restart <service>"
   

Podman/Container Issues

Check socket status:

ssh <proxmox-host> "pct exec <vmid> -- /run/current-system/sw/bin/systemctl status podman.socket"

List running containers:

ssh <proxmox-host> "pct exec <vmid> -- /run/current-system/sw/bin/podman ps -a"

SSH Connection Issues

If colmena fails with SSH errors:

1. Verify container is running on Proxmox
2. Check if SSH is listening: pct exec <vmid> -- /run/current-system/sw/bin/ss -tlnp | grep 22
3. Reboot container if necessary

DNS Server Deployment (Critical Infrastructure)

dns1 and dns2 are the backbone of the entire DNS infrastructure. NEVER deploy both simultaneously.

Sequential Deployment Procedure

Step 1: Deploy dns1

colmena apply --on dns1 --impure

Step 2: Verify dns1 is working

# Test DNS resolution through dns1
dig @dns1 google.com

# Check that dns1 responds correctly
ssh dns1 systemctl status blocky

Step 3: Only after dns1 is verified, deploy dns2

colmena apply --on dns2 --impure

Step 4: Verify dns2 is working

# Test DNS resolution through dns2
dig @dns2 google.com

# Check that dns2 responds correctly
ssh dns2 systemctl status blocky

DNS Verification Commands

# Quick DNS health check
dig @dns1 google.com +short
dig @dns2 google.com +short

# Check service status
ssh dns1 systemctl status blocky
ssh dns2 systemctl status blocky

# View recent logs
ssh dns1 journalctl -u blocky -n 20
ssh dns2 journalctl -u blocky -n 20

If DNS Fails After Deploy

1. Check if the service is running: ssh dns1 systemctl status blocky
2. Check logs for errors: ssh dns1 journalctl -u blocky -n 50
3. Verify network connectivity: ssh dns1 ping -c 3 8.8.8.8
4. If necessary, reboot: ssh dns1 sudo reboot

Common Colmena Patterns

Deploy All Gitea Runners

colmena apply --on gitea-runner-1,gitea-runner-2,gitea-runner-3 --impure

Deploy Monitoring Stack

colmena apply --on prometheus,grafana --impure

Update Secrets Before Deploy

just update-secrets
colmena apply --on <hostname> --impure

File Locations

Related Skills

• provision-nixos-server: Create new servers from scratch
• For creating new hosts, use /provision-nixos-server skill instead