Bug Report: v0.3.0 Fresh Install Fails — Migration 0071 DuplicateColumn on PostgreSQL

[plazmabokor]

Subject: Bug Report: v0.3.0 Fresh Install Fails — Migration 0071 DuplicateColumn on PostgreSQL

Hi,

I wanted to report a bug I encountered while doing a fresh install of Circuit Breaker v0.3.0 on a Raspberry Pi 3 (aarch64, Raspberry Pi OS Lite 64-bit / Debian 13 Bookworm).

Summary

Fresh native installs consistently fail to start the backend API due to a DuplicateColumn error in migration 0071_network_peer_connection_type. The backend enters a crash loop and never becomes healthy.

Environment

• Hardware: Raspberry Pi 3 (aarch64)
• OS: Raspberry Pi OS Lite 64-bit (Debian GNU/Linux 13 Bookworm)
• Install method: Native installer via --local-bundle flag
• Circuit Breaker version: v0.3.0

Steps to reproduce

1. Run a fresh native install using install.sh with the arm64 bundle
2. Installer completes successfully up to "Starting Application Services"
3. Backend API fails to start
4. Run: journalctl -u circuitbreaker-backend --no-pager -n 20

Error message

sqlalchemy.exc.ProgrammingError: (psycopg2.errors.DuplicateColumn) column "connection_type" of relation "network_peers" already exists
[SQL: ALTER TABLE network_peers ADD COLUMN connection_type VARCHAR]

Root cause analysis

Migration 0071_network_peer_connection_type.py attempts to add the connection_type column using batch_alter_table. Although the migration code checks for existing columns via inspector.get_columns(), this check does not work correctly with PostgreSQL in batch mode — the column check passes but the ALTER TABLE still fails with a DuplicateColumn error.

Because the failure happens inside a transaction, the entire migration is rolled back, including the alembic_version table entries. This means every restart triggers a full re-run of all migrations from scratch, hitting the same error every time — creating a permanent crash loop with no way to recover without manual intervention.

Additional notes

The release notes for v0.3.0 mention: "Fixed migration reruns throwing duplicate-table errors" — however this fix does not appear to cover this specific case on fresh PostgreSQL installs.

A simple fix would be to replace the batch_alter_table block in 0071 with direct PostgreSQL DDL using ADD COLUMN IF NOT EXISTS:

def upgrade() -> None:
    conn = op.get_bind()
    conn.execute(sa.text("ALTER TABLE network_peers ADD COLUMN IF NOT EXISTS connection_type VARCHAR"))
    conn.execute(sa.text("ALTER TABLE network_peers ADD COLUMN IF NOT EXISTS bandwidth_mbps INTEGER"))

This is idempotent and PostgreSQL-native, which avoids the batch_alter_table issue entirely.

Happy to provide any additional logs or information if needed. Thanks for the great project!

Best regards