c39465c2bd
Some checks failed
Test / test (push) Has been cancelled
Co-authored-by: Cursor <cursoragent@cursor.com>
205 lines
7.4 KiB
Markdown
205 lines
7.4 KiB
Markdown
# VM Placement Explanation - Why VMs Don't Need to Be on Both Servers
|
|
|
|
**Date:** 2025-11-27
|
|
**Question:** Why are VMs 100-103 required on both servers?
|
|
|
|
## Short Answer
|
|
|
|
**VMs 100-103 are NOT required on both servers.** They are deployed once and can run on either node in the Proxmox cluster. The cluster provides high availability through VM migration, not duplication.
|
|
|
|
## Architecture Overview
|
|
|
|
### Current Setup
|
|
- **Proxmox Cluster:** 2 nodes (ML110 and R630)
|
|
- **VMs 100-103:** Deployed on ML110 (can run on either node)
|
|
- **Shared Storage:** NFS (when configured) allows VM migration
|
|
|
|
### How It Works
|
|
|
|
```
|
|
┌─────────────────────────────────────────────────────────┐
|
|
│ Proxmox VE Cluster (hc-cluster) │
|
|
│ │
|
|
│ ┌──────────────┐ ┌──────────────┐ │
|
|
│ │ ML110 │◄───────►│ R630 │ │
|
|
│ │ (Node 1) │ Cluster │ (Node 2) │ │
|
|
│ │ │ Network │ │ │
|
|
│ └──────┬───────┘ └──────┬───────┘ │
|
|
│ │ │ │
|
|
│ └──────────┬─────────────┘ │
|
|
│ │ │
|
|
│ ┌─────▼─────┐ │
|
|
│ │ NFS │ │
|
|
│ │ Storage │ │
|
|
│ │ (Shared) │ │
|
|
│ └─────┬─────┘ │
|
|
│ │ │
|
|
│ ┌──────────┼──────────┐ │
|
|
│ │ │ │ │
|
|
│ ┌────▼───┐ ┌───▼───┐ ┌───▼───┐ ┌───▼───┐ │
|
|
│ │ VM 100 │ │VM 101 │ │VM 102 │ │VM 103 │ │
|
|
│ │ │ │ │ │ │ │ │ │
|
|
│ │ Can run│ │ Can │ │ Can │ │ Can │ │
|
|
│ │ on │ │ run on│ │ run on│ │ run on│ │
|
|
│ │ either │ │ either│ │ either│ │ either│ │
|
|
│ │ node │ │ node │ │ node │ │ node │ │
|
|
│ └────────┘ └───────┘ └───────┘ └───────┘ │
|
|
│ │
|
|
└─────────────────────────────────────────────────────────┘
|
|
```
|
|
|
|
## Key Concepts
|
|
|
|
### 1. Cluster = Shared Management, Not Duplication
|
|
|
|
A Proxmox cluster means:
|
|
- **Shared management:** Both nodes managed together
|
|
- **Shared storage:** VMs stored on shared storage (NFS)
|
|
- **VM migration:** VMs can move between nodes
|
|
- **High availability:** If one node fails, VMs can run on the other
|
|
|
|
**It does NOT mean:**
|
|
- ❌ Duplicate VMs on both nodes
|
|
- ❌ VMs running simultaneously on both nodes
|
|
- ❌ Separate VM instances per node
|
|
|
|
### 2. VM Placement Strategy
|
|
|
|
**Current Deployment:**
|
|
- VMs 100-103 are deployed on ML110
|
|
- They can be migrated to R630 if needed
|
|
- Only one instance of each VM exists
|
|
|
|
**Why Deploy on One Node Initially:**
|
|
- Simpler initial setup
|
|
- ML110 has SSH access configured
|
|
- Can migrate later if needed
|
|
|
|
**When to Migrate:**
|
|
- Load balancing (spread VMs across nodes)
|
|
- Maintenance (move VMs off node being maintained)
|
|
- Failure recovery (automatic or manual migration)
|
|
|
|
### 3. High Availability Options
|
|
|
|
#### Option A: Manual Migration (Current Setup)
|
|
- VMs run on one node
|
|
- Can manually migrate if node fails
|
|
- Requires shared storage (NFS)
|
|
|
|
#### Option B: HA Groups (Future)
|
|
- Configure HA groups in Proxmox
|
|
- Automatic failover if node fails
|
|
- Requires shared storage and quorum
|
|
|
|
#### Option C: Load Balancing
|
|
- Distribute VMs across both nodes
|
|
- Better resource utilization
|
|
- Still one instance per VM
|
|
|
|
## VM Details
|
|
|
|
### VM 100 - Cloudflare Tunnel
|
|
- **Current Location:** ML110
|
|
- **Can Run On:** Either node
|
|
- **Why:** Single instance sufficient, can migrate if needed
|
|
|
|
### VM 101 - K3s Master
|
|
- **Current Location:** ML110
|
|
- **Can Run On:** Either node
|
|
- **Why:** Single K3s master, can migrate if needed
|
|
|
|
### VM 102 - Git Server
|
|
- **Current Location:** ML110
|
|
- **Can Run On:** Either node
|
|
- **Why:** Single Git server, can migrate if needed
|
|
|
|
### VM 103 - Observability
|
|
- **Current Location:** ML110
|
|
- **Can Run On:** Either node
|
|
- **Why:** Single observability stack, can migrate if needed
|
|
|
|
## When You WOULD Need VMs on Both Servers
|
|
|
|
### Scenario 1: Separate Environments
|
|
- **Dev on ML110, Prod on R630**
|
|
- Different VM IDs (e.g., 100-103 on ML110, 200-203 on R630)
|
|
- Not a cluster, separate deployments
|
|
|
|
### Scenario 2: Load Balancing
|
|
- **VM 100, 102 on ML110**
|
|
- **VM 101, 103 on R630**
|
|
- Still one instance per VM, just distributed
|
|
|
|
### Scenario 3: High Availability Pairs
|
|
- **VM 100 primary on ML110, standby on R630**
|
|
- Requires application-level HA (not Proxmox)
|
|
- More complex setup
|
|
|
|
## Current Architecture Benefits
|
|
|
|
### ✅ Advantages of Current Setup
|
|
1. **Simplicity:** One deployment, easier management
|
|
2. **Resource Efficiency:** No duplicate resource usage
|
|
3. **Flexibility:** Can migrate VMs as needed
|
|
4. **Cost:** Lower resource requirements
|
|
|
|
### ⚠️ Considerations
|
|
1. **Single Point of Failure:** If ML110 fails, VMs need migration
|
|
2. **Load Distribution:** All VMs on one node may cause resource contention
|
|
3. **Maintenance:** Need to migrate VMs for ML110 maintenance
|
|
|
|
## Recommendations
|
|
|
|
### For Current Setup
|
|
- **Keep VMs on ML110** (where they are now)
|
|
- **Configure shared storage** (NFS) for migration capability
|
|
- **Test VM migration** between nodes
|
|
- **Monitor resource usage** on ML110
|
|
|
|
### For Future Optimization
|
|
- **Distribute VMs** across both nodes for load balancing:
|
|
- ML110: VM 100, 102
|
|
- R630: VM 101, 103
|
|
- **Configure HA groups** for automatic failover
|
|
- **Monitor and balance** resource usage
|
|
|
|
## Migration Example
|
|
|
|
### How to Migrate a VM
|
|
|
|
**Via Web UI:**
|
|
1. Select VM → Migrate
|
|
2. Choose target node (R630)
|
|
3. Start migration
|
|
|
|
**Via CLI:**
|
|
```bash
|
|
# Migrate VM 100 from ML110 to R630
|
|
qm migrate 100 r630 --online
|
|
```
|
|
|
|
**Via API:**
|
|
```bash
|
|
curl -k -X POST \
|
|
-H "Cookie: PVEAuthCookie=..." \
|
|
-H "CSRFPreventionToken: ..." \
|
|
-d "target=r630" \
|
|
"https://192.168.1.206:8006/api2/json/nodes/pve/qemu/100/migrate"
|
|
```
|
|
|
|
## Summary
|
|
|
|
**VMs 100-103 are NOT required on both servers.** They are:
|
|
- ✅ Deployed once (currently on ML110)
|
|
- ✅ Stored on shared storage (when NFS configured)
|
|
- ✅ Can run on either node in the cluster
|
|
- ✅ Can be migrated between nodes as needed
|
|
|
|
The cluster provides **high availability through migration**, not duplication. This is the standard Proxmox cluster architecture.
|
|
|
|
---
|
|
|
|
**If you need VMs on both servers for a specific reason, please clarify the requirement and we can adjust the architecture accordingly.**
|
|
|