docs: add EC2 infrastructure reference and resize recommendations

Documents current instance specs, volume history (25GB→100GB expansion),
disk usage breakdown, and recommended next resize path (t3.xlarge) with
rationale and procedure.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Author: sragss

.claude/infrastructure.md

@@ -0,0 +1,67 @@
+# EC2 Infrastructure
+
+## Current Setup
+- **Instance:** `i-0e2fb16c6e0fb83d0` (t3.large — 2 vCPU, 8 GB RAM)
+- **Volume:** `vol-0ce6db6bfe7959e24` (gp3, 100 GB)
+- **Elastic IP:** `100.31.229.192` (eipalloc-0155d2c49a8667e13)
+- **Security group:** `sg-083530ad37dce5694` (craigclaw-sg)
+- **Region:** us-east-1
+- **SSH:** `ssh -i ~/.ssh/craigclaw-key.pem ubuntu@100.31.229.192`
+
+## Volume History
+- Originally 25 GB (default), expanded to 100 GB on 2026-02-13
+- Snapshot `snap-02aef7fb07c1e218a` taken pre-resize as backup
+- Expansion was in-place (growpart + resize2fs), no reboot needed
+
+## Instance History
+- Started as t3.small (2 GB RAM). OOM incident 2026-02-11 — OpenClaw gateway
+  hit ~866 MB RSS, OOM killer looped, SSH starved. Resized to t3.large (8 GB).
+
+## Disk Usage (as of 2026-02-13, post-expansion)
+Major consumers: pnpm store (~6 GB), poncho + StableStudio + x402email repos
+with node_modules (~7.4 GB), npm cache (~1.3 GB). These are leftover from
+other projects and can be cleaned up if space gets tight again.
+
+## Recommended Next Resize: t3.xlarge
+
+When Craig needs more power, the recommended upgrade is **t3.xlarge**
+(4 vCPU, 16 GB RAM, ~$121/mo vs current ~$61/mo).
+
+**Why t3.xlarge:**
+- Craig's workload is bursty (agent gets task, builds, idles). Burstable
+  instances are designed for this — accumulate CPU credits while idle, spend
+  them during builds.
+- 4 vCPU doubles parallel build capacity.
+- 16 GB RAM gives headroom for OpenClaw gateway (~900 MB peak) + builds +
+  pnpm installs running simultaneously. No more OOM risk.
+- Zero-migration resize: stop instance, change type, start. ~1 min downtime.
+  Elastic IP stays attached.
+
+**Why not dedicated CPU (c7i/m7i):**
+- Dedicated cores matter for sustained CPU (CI servers, databases). Craig is
+  idle most of the time — paying for dedicated cores he won't use is waste.
+- c7i.xlarge is $130/mo for the same 4 vCPU / 8 GB but without burst credits.
+
+**Why not ARM/Graviton (c7g/m7g):**
+- ~20% cheaper per core, but requires full instance migration (new AMI, rebuild
+  environment, transfer secrets, re-test deploy pipeline). Hours of work to
+  save ~$15-20/mo.
+- Revisit if we ever rebuild the box from scratch.
+
+**Resize procedure:**
+```bash
+# 1. Stop instance
+aws ec2 stop-instances --instance-ids i-0e2fb16c6e0fb83d0 --region us-east-1
+aws ec2 wait instance-stopped --instance-ids i-0e2fb16c6e0fb83d0 --region us-east-1
+
+# 2. Change type
+aws ec2 modify-instance-attribute --instance-id i-0e2fb16c6e0fb83d0 \
+  --instance-type '{"Value":"t3.xlarge"}' --region us-east-1
+
+# 3. Start instance (Elastic IP re-associates automatically)
+aws ec2 start-instances --instance-ids i-0e2fb16c6e0fb83d0 --region us-east-1
+aws ec2 wait instance-running --instance-ids i-0e2fb16c6e0fb83d0 --region us-east-1
+
+# 4. Verify
+ssh -i ~/.ssh/craigclaw-key.pem ubuntu@100.31.229.192 "nproc && free -h"
+```