Ethereum Prometheus metrics mirror

GET /v1/eth/metrics/{host_id}/{layer} returns the raw Prometheus exposition text (text/plain; version=0.0.4) for a single Ethereum node in the platform's mesh, identified by host_id, for a single client layer. The platform scrapes the upstream client's metrics endpoint on your behalf, runs the payload through a label scrubber that removes operator-identifying data (instance IPs, hostnames, pod/node/namespace labels, and any raw IP-form label values), and serves the result. You can point Prometheus, Grafana Agent, or any scraper at this URL.

Responses are cached briefly (default 5 seconds) so that repeated scrapes don't hammer the upstream node. Every response carries an X-Cache-Status header (miss, fresh, or stale). If the upstream node is temporarily unreachable, the mirror serves the last-known-good payload from a stale cache (default ceiling 60 seconds) and marks it X-Cache-Status: stale so your tooling can alert.

Use this when you want to observe the health of a specific node you are routed through — block height, peer counts, sync status — without exposing the node's internal address space. The layer selects which client family's metrics port and path to scrape; the requested layer must match the host's actual client family, and the host must advertise the corresponding metrics capability.

Path parameters

The body is standard Prometheus exposition text. Note that the instance label has been rewritten from the upstream node's IP:port to the public host_id, and any operator-only labels (hostnames, pod/node/namespace, raw IP values) have been stripped.

Response headers

Errors are returned as a small JSON envelope: {"error":"<code>","code":<status>}.

See Errors for the shared error envelope and conventions.

JavaScript (fetch)

const res = await fetch(
  "https://api.triport.io/v1/eth/metrics/eth-el-01/el-geth",
  { headers: { Authorization: `Bearer ${process.env.TRIPORT_API_KEY}` } }
);


if (!res.ok) {
  const err = await res.json();
  throw new Error(`metrics mirror failed: ${err.error} (${err.code})`);
}


const cacheStatus = res.headers.get("X-Cache-Status"); // "miss" | "fresh" | "stale"
const metrics = await res.text();
console.log(cacheStatus, metrics);

TypeScript SDK (@triport/sdk)

import { Triport } from "@triport/sdk";


const client = new Triport({ apiKey: process.env.TRIPORT_API_KEY! });


const { body, cacheStatus } = await client.ethereum.metrics.mirror({
  hostId: "eth-el-01",
  layer: "el-geth",
});


console.log(cacheStatus, body);

Python (triport-sdk)

import os
from triport import Triport


client = Triport(api_key=os.environ["TRIPORT_API_KEY"])


result = client.ethereum.metrics.mirror(host_id="eth-el-01", layer="el-geth")
print(result.cache_status, result.body)

• Scraping. Because responses are cached for a few seconds, you can safely set your scrape interval as low as 5 seconds — within a cache window you'll get X-Cache-Status: fresh and the upstream node is not re-hit.
• Stale data. A stale cache status means the upstream node was unreachable on the last attempt; the payload may be up to the stale ceiling (default 60s) old. Treat sustained stale as a node-health signal.
• Label scrubbing is one-way. Internal addressing is removed before the payload leaves the platform; you will never see upstream IPs, hostnames, or cluster topology labels. The instance label always carries the public host_id.
• Layer ↔ host must agree. The layer you request must match the host's client family and the host must expose the corresponding metrics capability; otherwise you get family_mismatch or capability_missing.
• Related: Ethereum JSON-RPC transport, Mesh health.