// Benutzung · Grundlagen
Checking whether a location is reachable
"Reachable" in Fleeger means fresh metrics are arriving from that location's spoke-agent — not that a network ping succeeds. It's derived from data, so it reflects the whole path actually working (agent running → NATS → ingestion).
How it's determined
ingestion-svc records a last_seen timestamp for each location every time a
metric message arrives (stored in db=ingestion, served at
/state/locations). The dashboard marks a location:
- reachable —
last_seenis less than 90 seconds ago - unreachable — no metrics in the last 90 seconds (or never seen)
An agent sends a heartbeat every poll interval (30s by default) and at least
every 30s, on a timer of its own, so a healthy location refreshes well inside
the 90s window however long a collection cycle takes. Each resource's
own data still arrives once per cycle, and its resource_unreachable alert
uses the same 90s default window, so keep the poll interval (configurable per
agent via FLEEGER_POLL_INTERVAL_SECS, or poll_interval_secs in the TOML
config) well under 90s. 0 is ignored: the file's interval, else 30s,
applies. Each Locations card also shows the raw "last seen … ago".
In the UI
Open the Dashboard → Locations. Each card's badge is the answer; the "last seen" line tells you how stale.
From the command line
Query the state endpoint directly (via the gateway, with a token) or hit
ingestion-svc internally in dev:
curl -s http://localhost:7403/state/locations | jqEach entry has name, first_seen, last_seen. Compare last_seen to now.
Cross-checks in the UI
The Services page helps narrow down where the path is broken:
- Agents tab — expand the agent's row. If its live NATS connection is
shown (RTT, uptime, subscriptions) the agent is at least connected to the
broker, so a stale
last_seenpoints at collection/publishing rather than the network. No connection detail means it isn't connected at all. - Services tab — a log tail for each hub service (
ingestion-svc,gateway-svc, …), so you can confirm the hub side is healthy. Note agent logs are not shipped to the hub — read those on the remote host.
If a location is unreachable
Work the path from the location outward:
- Is the spoke-agent running at that location? (in dev, the host process / its config). A stopped agent is the most common cause.
- Can the agent reach
inventory-svcto list its resources, and the targets (Proxmox / iDRAC / k8s API)? Check the agent log. - Can it reach NATS, and are its credentials scoped to this location? A mis-scoped credential is broker-denied (see the NATS layout in the metrics pipeline).
- Is
ingestion-svcup and consuming? If other locations are fresh, the hub is fine and the problem is at the location.