hobbs-iac/docs/plans/2026-03-08-hf-pdocker-01-phase2-migration.md

hf-pdocker-01 Phase 2 — Service Migration Back

For Claude: REQUIRED SUB-SKILL: Use superpowers:executing-plans to implement this plan task-by-task.

Goal: Migrate all failover services back to hf-pdocker-01 from hf-pdocker-02 and bart.

Architecture: Stop services on failover nodes, start on hf-pdocker-01, update IP references and routing. Order matters due to dependencies.

Dependency Chain:

• Mosquitto must be up before Home Assistant and Frigate MQTT config update
• Matrix/Synapse must be up before AI Agents
• Traefik must be up before external routing works
• AdGuard is DNS — move carefully to minimize resolution downtime

---

Task 1: Migrate Mosquitto (bart → hf-pdocker-01)

MQTT broker — Home Assistant and Frigate depend on it.

1. Deploy Mosquitto on hf-pdocker-01 using compose from /mnt/docker-data/compose/bart/smarthome/ (extract mosquitto service) or create standalone compose
2. Verify MQTT is listening on hf-pdocker-01:1883
3. Update Frigate config MQTT host back to 192.168.86.192 (localhost)
4. Restart Frigate to pick up new MQTT
5. Stop Mosquitto on bart

Task 2: Migrate Home Assistant + ESPHome (bart → hf-pdocker-01)

Smart home services — depend on Mosquitto.

1. Deploy HA + ESPHome on hf-pdocker-01
2. Update HA config: MQTT broker IP back to 192.168.86.192
3. Verify HA web UI on :8123, ESPHome on :6052
4. Stop HA + ESPHome on bart

Task 3: Migrate Matrix/Synapse (bart → hf-pdocker-01)

Matrix homeserver — AI agents and alertbot depend on it.

1. Deploy Synapse + synapse-db + wellknown on hf-pdocker-01
2. Verify Matrix federation and client API on :8008
3. Update matrix-alertmanager-receiver config to point to new Synapse
4. Stop Synapse stack on bart

Task 4: Migrate Guacamole (bart → hf-pdocker-01)

Remote desktop gateway — has local MySQL DB on bart (NFS lock workaround).

1. Copy guacamole-db data from bart:/var/lib/guacamole-db to hf-pdocker-01 local storage
2. Deploy Guacamole + guacd + guacamole-db on hf-pdocker-01
3. Verify web UI on :8199
4. Stop Guacamole stack on bart

Task 5: Migrate AI Agents (bart → hf-pdocker-01)

margaret, spirit, bob, clippy — depend on Matrix/Synapse.

1. Build margaret-openclaw:latest image on hf-pdocker-01
2. Update agent configs: Matrix homeserver URL back to 192.168.86.192:8008
3. Rename Matrix account state dirs
4. Deploy all 4 agents on hf-pdocker-01
5. Verify Matrix connectivity
6. Stop agents on bart

Task 6: Migrate AdGuard Home (hf-pdocker-02 → hf-pdocker-01)

DNS server — brief downtime expected during switchover.

1. Deploy AdGuard on hf-pdocker-01 using NAS config
2. Verify DNS resolution on :53 and web UI on :3000
3. Stop AdGuard on hf-pdocker-02

Task 7: Migrate Traefik (hf-pdocker-02 → hf-pdocker-01)

Reverse proxy — handles all external HTTPS routing.

1. Deploy Traefik on hf-pdocker-01 with NAS certs and dynamic configs
2. Update dynamic configs for new service IPs (services now on .192 instead of .167)
3. Verify HTTPS routing for mess/solo/yoda/vault/git.hobbs.farm
4. Stop Traefik on hf-pdocker-02

Task 8: Migrate Monitoring Stack (hf-pdocker-02 → hf-pdocker-01)

Grafana, Prometheus, Alertmanager, InfluxDB, Blackbox, matrix-receiver.

1. Deploy full monitoring stack on hf-pdocker-01
2. Update Prometheus config with correct service IPs (most services now back on .192)
3. Update Telegraf configs on all nodes to point InfluxDB to .192
4. Verify Grafana on :3001, Prometheus on :9095
5. Stop monitoring stack on hf-pdocker-02

Task 9: Migrate Farm Manager API (new deploy on hf-pdocker-01)

Was never failed over — needs fresh deploy.

1. Clone farm-manager from Gitea, build server image
2. Deploy Farm Manager API on :8888
3. Update config.json with correct node IPs
4. Verify dashboard and API

Task 10: Cleanup + Final Verification

1. Remove stopped failover containers on bart and hf-pdocker-02
2. Update all Prometheus blackbox targets for final service locations
3. Hot-reload Prometheus
4. Update Traefik dynamic configs
5. Verify all services healthy
6. Update CLAUDE.md, MEMORY.md, wiki