diff --git a/graph/decisions/proxmox-upgrade-stagger-guest-startup-to-avoid-io-storm-b25708.md b/graph/decisions/proxmox-upgrade-stagger-guest-startup-to-avoid-io-storm-b25708.md
new file mode 100644
index 00000000000..1c9d3e63c70
--- /dev/null
+++ b/graph/decisions/proxmox-upgrade-stagger-guest-startup-to-avoid-io-storm-b25708.md
@@ -0,0 +1,21 @@
+---
+id: b25708d0-f025-4d8e-8a68-69cfe8f604f1
+type: decision
+title: "Proxmox upgrade: stagger guest startup to avoid I/O storm"
+tags: [proxmox, homelab, performance, upgrade]
+importance: 0.7
+confidence: 0.8
+created: "2026-02-19T17:05:45.547388+00:00"
+updated: "2026-02-19T17:05:45.547388+00:00"
+---
+
+## Decision: Stagger Guest Startup After Proxmox Upgrades
+
+When starting all Proxmox guests after a reboot/upgrade, starting them all simultaneously causes massive I/O delay (50%+ for several minutes) because all Docker containers and services compete for the same LVM thin pool.
+
+Recommendations for Phase 2 and future reboots:
+1. Add sleep 30 between dependency groups during startup
+2. Disable onboot=1 on all guests before upgrade, re-enable after validation
+3. Restart Pi-hole Docker containers proactively after boot (UDP DNS may fail)
+4. Switch local desktop DNS to public resolvers (1.1.1.1) before shutting down guests
+5. Set DHCP reservation for Home Assistant VM to prevent IP changes