Your node's vital signs, straight off the wire.

Pearl Metrics renders a pearld Prometheus scrape the way an operator actually reads it: chain-tip freshness, peer health, wire traffic by command, per-method RPC latency from the real histogram buckets, mempool pressure, and per-second rates between scrapes. Everything is parsed in this page — nothing is uploaded anywhere.

Status of the endpoint — read first. The /metrics endpoint comes from upstream pearl-research-labs/pearl PR #310 (--metricslisten, default port 9105), which is not yet in a released pearld as of 2026-10-06. This dashboard is built against that PR's published metric catalogue (node/metrics/README.md on the branch) so it is ready the day the PR lands — until then, use it with a pearld built from that branch, or explore it with the clearly-labelled sample scrape. The endpoint is unauthenticated: upstream warns to bind it to loopback or a firewalled private address, never the public internet.

Get a scrape

Metrics endpoint URL

Auto-refresh every 15s

Enable it on the node with pearld --metricslisten=127.0.0.1:9105. Browsers usually can't fetch it from this HTTPS page (pearld sends no CORS headers) — running this page locally against a loopback node works, and paste always works.

…or paste a scrape (curl -s 127.0.0.1:9105/metrics)

No scrape loaded yet.

Health notes (heuristics, not consensus)

Chain

Peers

Mempool

RPC by method

Latency is computed from the pearld_rpc_request_duration_seconds histogram: mean is exact (sum ÷ count); p95 is approximated by interpolation inside the bucket where the rank falls, and is hidden when the rank falls in the +Inf bucket.

Method	Requests	Errors	Error %	Mean	p95 ≈

Wire traffic by command

Command	Messages in	Messages out	Bytes in	Bytes out

All series

Filter

Series	Type	Value

Includes the standard go_* and process_* series pearld exports on the same endpoint — they're listed, just not summarised above.

How this differs from the Node console

The Pearl Node console polls JSON-RPC (getinfo, getpeerinfo, …) for live state. This dashboard reads the Prometheus endpoint instead, which carries what RPC doesn't publish: cumulative RPC request/latency statistics per method, auth failures, wire bytes per P2P command, peer bans and rejection reasons, and block connect/disconnect counters. Operators running Prometheus + Grafana don't need this page at all — it's for a quick, dependency-free look at one scrape, on any machine, with nothing installed.