relay: bind the host LAN IP, admit the podman bridge subnet only

169.254.1.2 is not a local address on this host (netavark runs its
bridges in aardvark-dns' netns), so the relay cannot bind it — and a
loopback bind refuses container traffic. Probe-socket proof: a host
socket on 192.168.0.143:18094 answers the container's 169.254.1.2:18094
dial while the LAN IP refuses from the container. The relay binds
192.168.0.143:18093 and the deployer installs an iptables INPUT pair
(ACCEPT 10.89.0.0/16, DROP rest) so the unauthenticated endpoint stays
off the LAN.

serverAppConfigs-docs/Infrastructure/Alertmanager.md

@@ -19,7 +19,7 @@ Alertmanager receives firing rules from Prometheus, groups and dedupes them, and
 
 ## Why the webhook address is `host.containers.internal`
 
-The Signal relay is a host-side process, and the address it must bind is the one detail this design gets wrong once: **127.0.0.1 does not work**. From a bridge container, loopback is the container's own, and netavark DNATs *published* ports only — the LAN IP and the podman gateway are both refused for an unpublished host bind. `host.containers.internal` (169.254.1.2) is the one address that reaches a host-local listener from a bridge container, so the relay binds *that* address (measured 2026-09-17: a loopback bind refused the first webhook with `dial tcp 169.254.1.2:18093: connection refused`). 169.254.1.2 is link-local to the podman bridge — not on loopback, not on any LAN interface, reachable only from this host's containers — so the unauthenticated webhook endpoint stays off the wire.
+The Signal relay is a host-side process, and the address it must bind is the one detail this design gets wrong twice (2026-09-17). From a bridge container, loopback is the container's own, and netavark DNATs *published* ports only — so a relay bound to 127.0.0.1 refused the first webhook with `dial tcp 169.254.1.2:18093: connection refused`. The obvious fix — bind `host.containers.internal` itself — fails too: on this host netavark runs its bridges in aardvark-dns' netns, so 169.254.1.2 is not a local address in the host's namespace (`OSError: [Errno 99] Cannot assign requested address`). What works: bind the **host's LAN IP** (192.168.0.143) — netavark delivers the container's `169.254.1.2` dial to that address (proven with a probe socket: container → `169.254.1.2:18094` → host socket bound to `192.168.0.143:18094`, while the LAN IP refused from the container, the usual netavark quirk). Since that puts an unauthenticated endpoint on a host interface, the deployer installs an iptables INPUT pair admitting 18093/tcp from the podman bridge subnet (10.89.0.0/16) only — the rest of the LAN gets a DROP.
 
 ## The Signal relay
 
@@ -36,7 +36,7 @@ The Signal relay is a host-side process, and the address it must bind is the one
 ./scripts/deploy.sh dankclaw --service alertmanager
 ```
 
-`deploy_observability` installs the quadlet unit, `copy_alertmanager_config` installs the config, and `install_signal_relay` copies the user unit, enables and restarts it, then waits on `169.254.1.2:18093/health` before the container's own `/-/healthy` wait. The planner selects `alertmanager` for its quadlet file, `shared/prometheus/alertmanager.yml`, and the two relay files; `prometheus` ships the `alerting:` block that points at `alertmanager:9093`.
+`deploy_observability` installs the quadlet unit, `copy_alertmanager_config` installs the config, and `install_signal_relay` copies the user unit, installs the iptables pair, enables and restarts the relay, then waits on `192.168.0.143:18093/health` before the container's own `/-/healthy` wait. The planner selects `alertmanager` for its quadlet file, `shared/prometheus/alertmanager.yml`, and the two relay files; `prometheus` ships the `alerting:` block that points at `alertmanager:9093`.
 
 ## Gotchas
 

serverAppConfigs-docs/Infrastructure/Observability.md

@@ -40,7 +40,7 @@ Recorded 2026-09-11, on the day the stack moved to Quadlet units. Everything in
 
 - **Grafana has a datasource and one dashboard as of 2026-09-16; before that it had neither.** `shared/grafana/provisioning/` is installed to `/opt/grafana/provisioning` by the deployer and read by the container: the Prometheus datasource (`http://prometheus:9090`, fixed uid `prometheus`) and the **LLM fleet — capacity and reuse** dashboard (speculative acceptance, queue depth, prefix-cache reuse, decode step rate, throughput) covering the llama.cpp endpoints. **Re-cut on 2026-09-16**: the panels were keyed to `vllm:` series, which stopped existing when frank-medium moved to llama.cpp, and two of them (KV-pool fill, request-latency p95) have no llama.cpp equivalent at all — those two are retired rather than restated, and the acceptance panel replaces the KV one. A renamed query against a series that does not exist is a panel that draws nothing forever. Verified through the API rather than by eye — the datasource answers `llamacpp:` counters for `frank-medium`, and `GET /api/search?type=dash-db` returns uid `llm-fleet`. One trap worth knowing: a provisioned dashboard does **not** appear in the legacy `dashboard` table on this Grafana version (it lives in unified storage …
 - **Grafana's dashboard now also surfaces firing alerts.** The **Alerts firing (ALERTS)** panel graphs `ALERTS{alertstate="firing"}` for the LLM rules, because this stack has no other place where an alert becomes visible without opening Prometheus.
-- **Prometheus now evaluates rules, and they are delivered (2026-09-17).** `shared/prometheus/rules/llm-fleet.yml` ships with the config (the deployer's `copy_prometheus_config` installs both — a rule file nothing loads evaluates nothing, and Prometheus reports no error for that) and holds five rules for the LLM fleet: request queueing, prefix-cache collapse, speculative decoding off, endpoint down, plus the `job:llm_prefix_cache_reuse:rate5m` recording rule the alert and the dashboard share. **Two rules from the vLLM era were retired, not renamed**, because the series they watched no longer exist: KV-pool saturation (llama.cpp exports no KV-fill gauge; its consequence *is* queueing, which `LlmRequestsQueueing` already covers) and p95 request latency (no latency histogram). `LlmSpecDecodeOff` takes the second slot — the failure a client cannot see, since a lost drafter just halves the speed. Verified with `promtool check rules` and `/api/v1/rules` (`health=ok` for all five). **Delivery went live 2026-09-17**: the `alerting:` block in `prometheus.yml` points at `alertmanager:9093` (container name on `dankclaw_default` — the published port is loopback-only and would be wrong from inside Prometheus). Alertmanager groups by `alertname` (30 s group wait, 5 m interval, 4 h repeat) and webhooks the host-side **Signal relay** at `host.containers.internal:18093/alertmanager` — the relay binds 169.254.1.2, the host-gateway endpoint: a loopback bind refuses container traffic (measured on the first deploy), and a host-interface bind would put an unauthenticated endpoint on the LAN. The relay (`scripts/signal-relay.py`, a systemd *user* unit, not a quadlet — it runs python3 from the checkout and can be enabled) dedupes per group for 15 minutes and sends through **signal-cli account 318311** (`/usr/local/bin/signal-cli`, state in `~/.local/share/signal-cli`), Frank's existing account. The Grafana dashboard's ALERTS panel and Prometheus' `/alerts` remain the second and third surfaces.
+- **Prometheus now evaluates rules, and they are delivered (2026-09-17).** `shared/prometheus/rules/llm-fleet.yml` ships with the config (the deployer's `copy_prometheus_config` installs both — a rule file nothing loads evaluates nothing, and Prometheus reports no error for that) and holds five rules for the LLM fleet: request queueing, prefix-cache collapse, speculative decoding off, endpoint down, plus the `job:llm_prefix_cache_reuse:rate5m` recording rule the alert and the dashboard share. **Two rules from the vLLM era were retired, not renamed**, because the series they watched no longer exist: KV-pool saturation (llama.cpp exports no KV-fill gauge; its consequence *is* queueing, which `LlmRequestsQueueing` already covers) and p95 request latency (no latency histogram). `LlmSpecDecodeOff` takes the second slot — the failure a client cannot see, since a lost drafter just halves the speed. Verified with `promtool check rules` and `/api/v1/rules` (`health=ok` for all five). **Delivery went live 2026-09-17**: the `alerting:` block in `prometheus.yml` points at `alertmanager:9093` (container name on `dankclaw_default` — the published port is loopback-only and would be wrong from inside Prometheus). Alertmanager groups by `alertname` (30 s group wait, 5 m interval, 4 h repeat) and webhooks the host-side **Signal relay** at `host.containers.internal:18093/alertmanager` — the relay binds the host's LAN IP (192.168.0.143:18093, the one address netavark delivers that dial to), with an iptables pair admitting the podman bridge subnet only. The relay (`scripts/signal-relay.py`, a systemd *user* unit, not a quadlet — it runs python3 from the checkout and can be enabled) dedupes per group for 15 minutes and sends through **signal-cli account 318311** (`/usr/local/bin/signal-cli`, state in `~/.local/share/signal-cli`), Frank's existing account. The Grafana dashboard's ALERTS panel and Prometheus' `/alerts` remain the second and third surfaces.
 - **Loki is configured but unread.** Storage is filesystem with 30-day retention, `auth_enabled: false`, and the push path works; Grafana can now reach it in Explore, but no dashboard or alert consumes it.
 - **node-exporter runs its default collector set** — no collector tuning, no textfile collector for custom metrics.
 - **Alloy is the one configured piece.** It ships the host journal and podman container logs to Loki with `job` labels, and that pipeline is live.