Prometheus monitoring with Topologix

Monitoring & Observability

Connect any Prometheus-compatible HTTP API — Prometheus itself, VictoriaMetrics, Thanos, Mimir — and chart the result of any PromQL query on a node, or run an All-Services latency alarm across every service a query returns.

Requirements

• Network reach from the Topologix server (or a proxy agent) to the Prometheus HTTP API port (default 9090).
• No credentials — Topologix does not send Basic auth to Prometheus. If your instance sits behind auth, front it with a reverse proxy that strips the header or route through a Topologix proxy agent on the same network.

1 · Set up the integration

1. Open the Integrations pageClick Integrations in the toolbar, then find Prometheus in the catalogue (the search box narrows it instantly).
2. Enter the base URLPoint it at the Prometheus HTTP API — typically http://prometheus:9090. Topologix appends /api/v1/query internally, so do not include the /api/v1 suffix.
3. Pick a default proxy (optional)If the server can’t reach Prometheus directly, choose a proxy agent running inside that network. Each metric source can override this later.
4. Create, then verifyClick Create, then reopen the integration and click Test on the configured card to verify connectivity.

2 · Add it to a flow

1. Select a nodeOpen your flow, click the node that should carry the metric, and open the Metrics tab in the Configure panel.
2. Add a sourceClick + Add, choose your Prometheus integration, then pick a mode: Preset (CPU %, Memory %, Error Rate, Latency p99), Custom PromQL for any query, or All Services Alarm to watch every service a query returns.
3. Scope and saveOptionally scope to specific jobs, set the polling interval (10s floor), add a threshold, and click Add Source. The first value lands within a second; the sparkline builds as history accumulates.

3 · Beyond presets: custom queries & the All-Services alarm

Custom PromQL

Chart the result of any PromQL query that returns an instant vector — the first series’ value lands on the card each poll.

• Scope custom queries inside the PromQL itself (e.g. {job="api"}) — the job picker applies to presets.
• Tick Treat empty result as 0 for rate/count queries that are legitimately empty when idle; leave it off so broken queries and disappearing metrics surface as errors.

All Services Alarm

One source that watches the whole fleet. A fleet query counts your services (default count(group by (job)(up))), and each section — CPU %, Memory %, Error Rate and Latency p99 come pre-configured — runs its own per-job query with its own threshold.

• Every service a section returns gets independent breach state: a fresh breach raises one alert for that service and section; acknowledging mutes it without hiding it.
• Breach state auto-clears after sustained clean polls (default 10), and services that disappear from results age out — no ghost rows.
• The card shows “N services · M breaching” with a per-section status row; one broken section query never blanks the card.