Problem

The supervisor port configuration was incorrectly changed from 9000 to 8080, breaking agent deployments.

Symptom:

{"error_message":"No instances with available capacity"}

Root Cause:

1. Original config had PORT=9000 (wrong variable name, but worked because code defaults to 9000)
2. Recent fix changed it to SUPERVISOR_PORT=8080 (right variable, wrong port)
3. Security group allows port 9000, not 8080
4. PAC reaches supervisor via private IP on port 9000 (not through ALB)
5. Restart killed all running agents

Solution

Standardize on SUPERVISOR_PORT=9000 everywhere:

Phase 1: Immediate EC2 Fix

• Update /etc/par-supervisor.conf: SUPERVISOR_PORT=9000
• Restart supervisor service
• Verify supervisor responds on port 9000

Phase 2: Fix Deploy Script

Change scripts/deploy_ec2_par.sh to use port 9000:

• Line 239: Set SUPERVISOR_PORT=9000 (not 8080)
• Lines 337, 340, 343, 356, 380, 383, 386, 394: Change all :8080 → :9000

Phase 3: Consistency Check

All other files already correct:

• ✅ docs/SUPERVISOR_README.md: States port 9000
• ✅ systemd/pixell-supervisor.service: Has SUPERVISOR_PORT=9000
• ✅ src/pixell_runtime/supervisor/__main__.py: Defaults to 9000

Impact

• Fixes broken agent deployments
• Restores supervisor to working state
• Standardizes port configuration across all files

Testing

1. Deploy fixed script to EC2
2. Verify supervisor responds: curl http://localhost:9000/health
3. Redeploy agents (PAF-Core, vivid-commenter)
4. Verify agents deploy successfully