# System Health

## What this page is for

A live dashboard of "is each platform service responding right now." An on-call engineer opens it during an incident to find which service is the source of the trouble; an operator opens it during a deploy to confirm everything came back up.

## Sidebar location & role gating

* **Sidebar group**: Settings → Operations.
* **Path**: `/system-health`.
* **Keyboard hint**: `G H`.
* **Minimum role**: any authenticated role. The underlying `/system/health` endpoint is intentionally low-friction so even a `VIEWER` can confirm a healthy stack.

## What you see

* **Overall status banner** — top of page. Green if every service is healthy, amber if any service is degraded, red if any service is unhealthy.
* **Per-service tiles** — one tile per backend service with: name, status badge, last-success timestamp, response-time p95, the route checked.
* **Aggregate latency block** — p50, p95, p99 across the whole platform.
* **Active alerts** — any Prometheus alerts currently firing. Each alert links to the relevant runbook.
* **Kill switch indicator** — chip at the top showing whether the kill switch is currently engaged.

## Backend calls

| Action           | HTTP | API path         | Service                                  |
| ---------------- | ---- | ---------------- | ---------------------------------------- |
| Aggregate health | GET  | `/system/health` | gateway (deep probe of every downstream) |

The gateway's `/system/health` consults every downstream service and aggregates the results into one response. The UI never calls the per-service healths directly.

## Auto-refresh & realtime

* **30-second poll** via `setInterval(() => fetchHealth(), 30_000)` at `ui/src/pages/SystemHealth.jsx:158`.
* **No SSE.** Health checks are cheap; polling is fine.

## Per-agent scoping

No. Platform-wide health is tenant-agnostic.

## Empty states

The page does not have a meaningful empty state — `/system/health` always returns a response. A failed fetch shows "System health check unavailable" as a banner.

## Edge cases & known gotchas

* **All services green but customers complaining**: the health check is a deep probe but not an end-to-end probe. A slow downstream tool (Anthropic, Groq) won't show here. Cross-reference Settings → Observability for latency.
* **One service amber for over an hour**: degraded mode kicked in for that service. The audit chain still records every consult with `service_status="degraded"` so behavior is observable.
* **`/status` vs `/system/health`**: `/status` is the public customer-visible page and only carries gateway-internal latency. `/system/health` is the operator view with downstream RTTs. Don't confuse the two.
* **Per-EC2 flap**: `/system/health` is gateway-internal so the two EC2s see the same RDS / Redis state. Disagreement between the two would point at a Docker DNS issue on one host.

## Related docs

* [Gateway service](/services/hot-path/gateway.md)
* [Deployment Topology](/architecture/deployment-topology.md)
* [Observability UI](/ui/settings-sub-pages/observability.md)

## Screenshot


---

# Agent Instructions: Querying This Documentation

If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter:

```
GET https://docs.aegisagent.in/ui/settings-sub-pages/system-health.md?ask=<question>
```

The question should be specific, self-contained, and written in natural language.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.